rman booklet

8/9/2019 Rman Booklet

1/120

An Introduction to

3D Computer Graphics

Version 6.0 PDF 1995

Exploring Photo-Realismwith MacRenderMan

Malcolm A. Kesson


2/120

CONTENTS

1 IntroductionInteractivity v scripting

Illusions and interfaces

RenderMan

What is a script?

Why u se scripting?

What s the catch?

2 Getting Started

Overview

Using a d efault camera

Setting a p erspective view

Rotating an object

Scaling

Assembling an object

Syntax and the structu re of a RIB file

3

Transformations

Translation

Rotation

Scaling

Skewing

Applying transformations

App lying sequences of transformations

4 Shading the basics

Using lights and materialsApp lying an image to an object

Preparing an image for texture map ping

Using an im age to d isplace a su rface

Avoiding rendering errors and improving performance

5 Shaping Up Library Objects and Polygons

Overview quadrics and p olygons

RenderMans library of quadric surfaces

Placing objects in the w orld


3/120

Positioning the w orld relative to the camera

Modelling a coffee mu g

The effects of scaling and tran slation

Reusable geometry

Playing w ith materials surface shaders w ood, carpet and spatterMaking a comp osition the wron g way

Making a comp osition th e correct w ay

Another w ay of group ing objects

Sum mary of method s relating to th e group ing of objects

A simp le polygon mod el (to be added )

6 An Improved Camera

Overview

Depth of fieldMotion blur

Field of view

Matching a VR camera to a real camera

7 Animation

Using FrameUP

Animated texture and displacement maps

8 Basic Lighting

Overview

Defining a light source

Types of light sou rces descriptions

Types of light sou rces examp les

An examp le script

Reference

Positioning lights in space

9 Advanced lighting Casting Shadow s

Overview

An examp le script

The shad ow algorithm: how it works

An examp le animation


4/120

10 Importing Fragments

Overview

A sample fragment

Importing correctly

Importing incorrectlyFragments an d objects

Restrictions

Appendix A Overview of MacRenderMan

Appendix B RenderMan Quick Reference

Appendix C Shaders Reference

Appendix D Projects

Separating Shape from Shad ing

Combining the real and the imaginary

Three Dimensional Icons for a Graph ical User Interface


5/120

These notes are intend ed to explain the basics of theRend erMan system by

provid ing a series of examp les of its use in theMacintosh environm ent.

Although a nu mber of exercises and projects have been included they will

only be effective when u sed by those who wish to explore and experiment

with the RenderMan system.

I wish to acknowledge the support I received from PIXAR, especially in

graciously providing several pre-release versions of their p hoto-realisticrenderer that was being ported to the new RISC based Power-Macintosh

compu ters at the time that I was p reparing th is booklet for teaching und er-

grad uate stud ents of graphic design the principles of 3D comp uter grap hics.

Malcolm Kesson

April 1994

Wellington

New Zealand

Preface


6/120

In Progress

The following sections are incomp lete,

5 Shaping Up

A sub-section dealing with the w ay in w hich RenderMan hand les polygons

has yet to be added. Several modelling exercises using polygons will also

form par t of this chapter.

10 Importing Fragments

The contents of this chap ter are almost complete but som e diagram s have

yet to be included and the body copy still requires some editing.

Add itional sections that may be add ed later includ e a general explanation

about viewing an d shad ing as well as the following,

11 Advanced Texture Mapping

A chapter dealing w ith the r elationships between cartesian space, texture

space and par ameter space. Use of texture s and t par ameters to control

the texturing of polygons and the use of the comm and TextureCoordinates

to likewise control the texturing of quadric surfaces.

12 Solid Modelling

A chapter d ealing w ith the p rinciples of boolean operations on sets of

enclosed objects.

Malcolm Kesson

Apr il 9th 1995


7/120

Most designers, especially those who are new to compu ter systems, assum e

the only way to work w ith a compu ter is to use interactive software. Indeed,

grap hical user interfaces (GUIs, pronou nced goo-eez) are taken so mu ch forgranted that it may appear strange, if not bizarre, to reject the ease-of-use that

such systems offer in favour of an environmen t based on text an d scripting.

What possible advan tage could th ere be in using a keyboard rather than a

mou se for grap hical inpu t? Why exchange p ull-dow n m enus, floating

wind ows, dialog boxes and icons for an u nfamiliar w ay of making images that

requires a large investment of time to m aster and th at emp hasises though t,

care and perfect attention to detail? The answer to these questions lies

principally in th e natu re of a GUI.

The problem w ith interactive software is that their interfaces are d esigned to

hide the intricacies of the algorithms and techniques u pon wh ich they arebased. Infact, just as a conjurer deceptively presen ts fiction as fact, GUIs

organise their illusions aroun d metaph ors that rou tinely entice us to accept

the impossible. For example, in illustration software such as Aldus FreeHand

or Adobe Illustrator, users interact with elements of their artw ork as if they are

on separate layers. Even op erating systems encourage u sers to perceive

wind ows as being stacked and ordered into layers. Thus, window s can be

moved to the front or sent to the back. But th e notion that an image on a

compu ter screen can have d epth, let alone be comp rised of layers, is pu re

fiction. This course is intended to take you behind the illusions in order to

more fully understand the principles of 3D mod elling and rendering.

Working in the area of 3D compu ter grap hics without a GUI involves

communicating directly with a software package called a renderer. A renderer

is somew hat like a laser printer but instead of turning a 2D page description,

norm ally in a comp uter langu age called PostScript, into a p rinted image, it

accepts a 3D scene description and converts, or rend ers, it as an image that is

either viewed on the comp uter monitor, or saved as an image file. Because

most renderers are embedd ed w ithin an interactive modeller or animation

system the w ays in w hich they can be used are strictly limited by the host

software. Infact, the only peop le who can really get at the renderer are the

program mers wh o wrote the modelling or animation software!

Rend erers also form p art of software libraries used on high-end graph ics

workstations. But th ese require a kn owledge of a p rograming langu age such

as C, and traditionally, artists and designers have not been given access to

such skills. Fortunately, there is a renderer that su pp orts the typ e of comm un -

ication that we requ irePRMAN is part of the innovative RenderMan system

developed by PIXAR. RenderMan is intended to sup port the prod uction of

ph oto-realistic images based on a mini langu age called RIBRenderMan

Interface Bytestream. The intention of Rend erMan is to separate mod elling

from rend ering. In formulating th eir scene d escription stand ard, PIXAR

established a n um ber of rules by w hich th e characteristics of a virtual world ,

Introduction

Interactivity v scripting

Introduction 1 1

Illusions and interfaces

RenderMan


8/120

and a virtual camera to view th at world , can be commu nicated to a rend erer.

Because RenderMan organises the w ay m odellers can p ass information to

rend erers, PIXAR refers to their system as an interface. Information abou t a3D scene is written a s text and is stored in a RIB file. Norm ally these files are

prod uced by an interactive modelling or an imation app lication and are rarely

seen by a naive u ser of a comp uter system. How ever, because the d etails of

the RenderMan Interface have been published by PIXAR, anyone with access

to a w ord p rocessor can w rite or edit a RIB file by hand and can gain greater

control over th e entire image m aking p rocess. In this course you will use

Rend erMan to explore the fund amentals of ph oto-realistic 3D comp uter

graphics.

Scripts are used to convey information abou t a prod uction or performance.

The samples given below are examples of textural and symbolic scripts. Whatever form it takes, a script typ ically enables an au thor to pass su fficient

information about th e structure of a performan ce so that it can be, in some

sense, true or faithfull to the o riginal d esign. To work effectively, a script must

adh ere to certain rules that are understood by th e author an d th e performer.

For examp le, it wou ld be a d isaster for an actor playing the role of King H enry

to speak the lines given in italics, Aumer le locks the d oor.

Introduction 1 2

Enter Bolingbroke, crowned King Henry, with Harry Percy, and other nobles

AUMERLE (rising) Then give me leave that I may turn the key, 35 That no man enter till my tale be done.

KING HENRY Have thy desire.

Aumerle locks the door. The Duke of York knocks at the door and crieth

YORK (within) My liege, beware! Look to thyself! Thou hast a traitor in thy presence there.

King Henry draws his sword

KING HENRY (to Aumerle) Villain, Ill make thee safe.

What is a Script?

Richard II Act 5.3 scripting a theatrical performance

(NeXT Digital Press 1988)

The scripting you will use in this course is no different to an y other typ e of

traditional scriptingyou w ill be the au thor, PRMAN will be the p erformer

and you w ill both conform to the rules defined by Rend erMan.


9/120

If scripting is so powerful it is appropriate to ask why interactive software is so

pop ular? The answ er lies in the breadth and flexibility of modern software

design. In a produ ction environmen t the majority of tasks a designer needs to

add ress can be quickly and adequ ately tackled w ith interactive software. Butfor those who u nd ertake innovative and experimental wor k, scripting of one

kind or anoth er, can offer significant adv antages. At one end of the scale,

scripting can mean w riting an entire software p ackage and at the other end it

can mean w riting so-called macros for a spread sheet. In an edu cational

context, and m ore especially for a third level degree course, an investigative

app roach based on scripting means you will learn th e general principles of 3D

work rather than a single imp lementation. However, it should be recognized

that RIB scripts (files) are N OT norm ally wr itten by hand, but are u sually

prod uced by m odelling and animation software and these can hand le levels of

mod elling detail that wou ld be impossible for any hu man to reprod uce

manually.

The Labanotation System scripting human movement

low middle high

forward side backward

Why use scripting?

(The N ew Encyclopaed ia Britann ica vol 7 page 78)

From Three Pieces for String Quartet (No. 1) by Igor Stravinsky notation for scripting music

(The N ew Encyclopaed ia Britann ica vol 24 page 530)

Introduction 1 3


10/120

The RIB files in this section are intended to gu ide you throu gh th e basics of

working with Rend erMan. Each examp le has been carefully chosen to

introduce a broad selection of concepts relating to 3D computer graphics. Theexplanations accomp anying each examp le are quite brief and are only

intended to touch up on the ideas being presented. Dont worr y if the material

looks terribly confusing. As the course un folds, the p rinciples und erpinning

the concepts w ill be reiterated and illustrated m any times over.

When a technical term is used for th e first time it is printed in italics. You

should make every effort to un derstand its meaning before continuing w ith

the next section, Shap ing Up Library Objects and Polygon s.

At the conclusion o f this section you will be able to

write, save and send a simp le scene description to PRMAN,

set the basic characteristics of a virtual camera,

use the basic transformations ie. translation, rotation and scaling,

distinguish parameters from RIB statements,

differentiate world space from camera space,

un derstand the role of default settings.

Chapter headings:

Getting StartedShap ing Up Library Objects and Polygons

Camera

Digital Lighting

Digital Lighting and Shad ows

Shading

Getting Started

Overview

Getting Started 2 1


11/120

Getting Started 2 2

The pu rpose of this RIB file is to present a min imal scene to PRMAN and to

introdu ce the basics of interacting with the scripting an d rendering

environment.

The first two lines show the u se of the hash symbol # to ind icate these lines

are comm ents and m ust be ignored by the renderer. Comm ents can beincluded an yw here in a RIB file - they are the equ ivalent of post-it notices.

WorldBegin is a RIB statemen t and as such must be sp elled exactly as shown

ie. a single word with two capitalisations. Essentially it notifies Rend erMan

that objects comprising a scene descriptiona virtual worldare about to be

defined.

Disk is a RIB statement th at d efines, by the th reeparameters (num bers) that

follow it, a flat circular d isk situated 1 unit along the z axis, 0.5 units in rad ius

and a full 360 degrees in extent. App roximately, half the RIB statemen ts (or

command s as they w ill be referred to) you will use in this course requ ireparam eters. In all cases each p arameter m ust be separated by at least one

space. They m ay, however, be spread ov er several lines of text and have

comments at the end of each line, for examp le,

Disk1 #unit along the z axis0.5 #units in radius360 #degrees

Finally, WorldEnd indicates the description of the scene, or world, has been

completed. This small RIB file is interesting not just for w hat it describes bu t

also for w hat it om its. Althou gh it does not specify the characteristics of a

virtu al camera to view th e scene ie. its location and or ientation, or the surface

colour and ma terial characteristics of the d isk, or how it is lit, the rend erer is

able to prod uce an image because, in the absence of specific information, it

makes several assump tions and uses a num ber ofdefault settings. In

particular, RenderMan has p rovided

an orthographic view looking along the z axis with the camera and theworld sharing a common origin,

an image called Untitled measur ing 320x240 pixels,

a matte white surface for the disk that does not require external lighting.

Example 1

RIB

#ortho disk1.RIB#using a default cameraWorldBegin

Disk 1 0.5 360WorldEnd


12/120

Getting Started 2 3

#ortho disk1.RIB

#using a default camera

RIB

two comments about the scene, thefirst gives the name of the file and

the second is a brief note about thescene

x

y

z

y

x

z

WorldBegin

the default camera creates anorthographic view, 320x240 pixels,with a name supplied by RenderApp

begin describing the world

Disk 1 0.5 360

x

y

z

y

x

z

create a disk situated 1 unit alongthe z axis, 0.5 units in radius and360 degrees in circumference

description of the world completeWorldEnd

initially the origins of the camera

and world coincide

Visualizing example 1


13/120

Getting Started 2 4

Example 2

#perspective disk.RIB#setting a perspective viewProjection "perspective" "fov" 40WorldBegin

Translate 0 0 3Disk 0 0.5 360

WorldEnd

The purp ose of this file is to show the w ay in wh ich a virtu al camera u sing

perspective projection can be set up before the w orld is defined and also to

introdu ce the use oftranslation to m ove objects in a scene.

Before the w orld is d efined the statement Projection establishes a perspectiveview with afield of vision of 40 degrees - this is one of several statements that

control a virtual camera. Note that two of its three param eters are words. So

that RenderMan d oes not attempt to interp ret them as RIB statements or

command s, textual param eters are always given in qu otes ie. .

As in the p revious example, the scene consists of a single disk bu t this time

the origin of the coordinate system has been moved 3 units along the z axis

before the disk is defined. The Translate command has three parameters,

Translate x y z

to move, what m ay be thought of as a three dimensional cursor arou nd the

world space.

RIB


14/120

Getting Started 2 5

Projection "perspective" "fov" 40

RIB

x

y

z

y

x

z

WorldBegin

set the camera to give a perspectiveview with a field of vision of 40degrees, the size and name of theimage are supplied by RenderApp

begin describing the world

Disk 0 0.5 360

x

y

z

y

x

z

create a disk situated at the origin,0.5 units in radius and 360 degreesin circumference


initially the origins of the cameraand world coincide

Translate 0 0 3 move the origin 3 units along the zaxis

x

y

z

y

x

z



15/120

RIB

Getting Started 2 6

Example 3

#tube.RIB#rotating an objectProjection "perspective" "fov" 40WorldBegin

Translate 0 0 5Rotate -120 1 0 0Color 1 0 0Cylinder 1 -1 1 360

WorldEnd

In this scene a cylind er is introduced into a world space that has, for the

pu rpose of better viewing, been rotated and moved away from the camera.

The cylinder has also been coloured.

The RIB statement Cylinder, with four p arameters,

Cylinder radius depth height arc

shows how, like the disk in the previous example, an object from Render-

Mans library of pr imitive shap es can be u sed in a scene. The cylinder an d

disk, as well as the other su rfaces in the Rend erMan librar y, will be dealt w ith

in deta il in the next section.

This file uses ano ther type oftransformation,

Rotate angle x y z

which in this instance turns the coordinate system 120 degrees anti-clockwise

around the x axis BEFORE the or igin is translated 5 units along the z axis of

the world . Although th e renderer reads the transformations in the order in

wh ich th ey app ear, it postpones app lying them u ntil an object is declared, at

which time it back-tracks and uses the transformations from last to firstlike

bullets in the magazine of a gu n, the last one loaded is the first to be shot!

A cylinder is created within the redefined world coordinate system. Since the

camera is fixed to the old world origin, the renderer p rodu ces an image

looking slightly into the top of the cylinder. Using a fixed camera and trying

to obtain a p articular view ing angle by orientating an object in a scene is only

adequ ate for simple compositions. In the n ext section th e virtual camera is

positioned relative to the world - m uch like a hand held camera in real

photography.

Color (note the north American spelling) specifies a hue in terms of three

components,

Color red green blue

A colour is app lied to each ob ject un til another is declared in the RIB file.


16/120

x

y

z

yx

z

1

Getting Started 2 7


RIB

x

y

z

y

x

z

WorldBegin begin describing the world

Cylinder 1 -1 1 360

x

y

create a cylinder, 1 unit in radius,with a base 1 unit below and a top 1unit above the origin, 360 degrees incircumference



Translate 0 0 5Rotate -120 1 0 0

the transformations are applied inreverse order; first an anti-clockwiserotation of 120 degrees around the xaxis, followed by a translation of 5

units along the z axis

set the camera to give a perspectiveview with a field of vision of 40degrees, the size and name of theimage are supplied by RenderApp

y x

z

y

y x

z

2



17/120

Getting Started 2 8

Example 4

#scaled tube.RIB#scalingDisplay "scaling" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1WorldBegin

Scale 0.3 0.3 0.3Translate 0 0 5Rotate -120 1 0 0Cylinder 1 -1 1 360

WorldEnd

This example introdu ces the id ea ofscaling the w orld space, and th erefore,

any ob jects placed in it. It also illustrates the w ay in w hich the char acteristics

of a virtual camera can be further refined and controlled.

Like the p revious examp le, a cylind er is introduced into a world space that

has been rotated an d translated for better viewing. However, in th is examp le

the w orld sp ace has also been uniform ly redu ced to 30% of its original scale.

In this and all futu re scenes, the RIB statements Display an d Format are used

to provide ad ditional control over the imagery p rodu ced by th e virtual

camera. Display uses three p arameters to sp ecify

the name of the image,

where to put the image, and what information the image should contain.

Format uses three numeric parameters

Format image width image height pixel ratio

Althou gh it ap pear s first, Scale only takes effect after the rota tion and

translation h ave been app lied - remember, transformations are app lied in

reverse order. The scale statement uses three parameters,

Scale x y z

to enlarge or redu ce a coordinate system along its x, y and z axes.

RIB


18/120

Getting Started 2 9

Display "scaling" "framebuffer" "rgb"

Projection "perspective" "fov" 40Format 200 150 1

RIB

x

y

z

y

x

z

WorldBegin begin describing the world

Cylinder 1 -1 1 360

x

y

create a cylinder, 1 unit in radius,with a base 1 unit below and a top 1

unit above the origin, 360 degrees incircumference



Scale 0.3 0.3 0.3Translate 0 0 5Rotate -120 1 0 0

the transformations are applied inreverse order; first an anti-clockwiserotation of 120 degrees around the xaxis, then a translation of 5 unitsalong the z axis, followed by theuniform scaling

set the camera to give a perspectiveview with a field of vision of 40degrees, set the size of the image to200x150 pixels storing rgb information

y x

z

x

y

z

yx

z

y

y x

z

x

y

y x

z

1 2

3


321


19/120

Getting Started 2 10

Example 5

#goblet.RIB#assembling an objectDisplay "goblet" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 150 200 1

WorldBeginScale 1 1 1 #change these to squash and stretch the gobletTranslate 0 0 5Rotate -120 1 0 0

Color 1.0 0.978 0.34 #goldCylinder 1 0 1.5 360 #containerDisk 0 1 360 #base of the containerCylinder 0.25 -1.5 0 360 #stemDisk -1.5 1 360 #base of the goblet

WorldEnd

RIB

In this example a nu mber of basic library shapes are assembled into a simp le

goblet.

Color app lies a un iform yellow hue to the entire goblet. Experiment w ith

alternative colour schemes by introdu cing add itional color statements

between each part of the goblet. In p articular, add another d isk positioned a

little below the rim of the goblet so that it ap pears to contain a coloured

liquid.

The Scale statement, in effect, does noth ing because it app lies a un iform

scaling factor of one. Change the scaling factor of each pa ram eter to see how

the goblet can be ind ividu ally squashed an d stretched in height, wid th and

dep th, for example,

Scale 1 2 1


20/120

Conclusion

By the time you finish the examp les in th is section and , no d oubt, completed

a few mod ifications of your ow n, you w ill have been introdu ced to m any

concepts, not only in 3D compu ter graph ics in general, but also in theabstract world of RenderMan. This section concludes with a brief review of

th e syntax of Rend erMan and an over-view of the structu re of a RIB file.

Twelve RIB statements or command s were u sed in Getting Started - by the

conclusion of the course you will have d ealt with ap proximately 35 of the

entire range of 96 RIB commands. In addition to the hash symbol, the

following statements:

Projection/ Display/ Format define a virtual camera,

WorldBegin/ WorldEnd relate to the concept of a virtual world ,

Translate/ Rotate/ Scale are examples of transformations, and finally

Disk/ Cylind er insert library objects/ surfaces into the world .

Incidentally, the word s statement and command are used interchangably.

RIB statements form p art a language recognised by th e rend erer. By hu man

standard s it is an impovished language, but n onetheless, it is in its own r ight

a complete system of commu nication. Some statements go together in p airs,

WorldBeginWorldEnd

and bracket, wh at are called blocks of RIB. Other statements have words

and / or num bers, called param eters, associated with them, for example,

Translate 0 0 5Projection "perspective" "fov" 40

that, in the m ajority of cases, provide essential information w ithout w hich the

statement makes n o sense.

At the beg inning of a RIB file only a virtu al camera exists, and therefore, all

statements relate to it and to nothing else, for examp le,

DisplayProjectionFormat(anything else that is appropriate)

As soon as th e renderer reads WorldBegin, the camera is frozen and all

subsequent statements effect the v irtual w orld, for example,

WorldBeginObjects etc

WorldEnd

and finally, WorldEnd marks the completion of the scene description.

Syntax

Structure of a RIB file

Getting Started 2 11


21/120

There are four basic method s of chan ging or m odifying 3D objects; they can

be repositioned, reorientated, resized or distorted in space. These alterations

to an object, called transformations, are carried ou t relative to the origin of thecoordinate system and are known as

translation moving rotation turn ing scaling stretching or squ ashing skewing shearing

The illustrations show th e effect of app lying th e transformations to tw o cones

that are p ositioned at the or igin an d a short distance along the xaxis. In each

case the zaxis is pointing into th e page.

Transformations

Transformations 3 1

Translation

Translate 3 1 0

Rotation

Rotate 45 0 0 1

Scaling

Scale 0.5 0.5 0.5

Notice how the cone shifts toward theorigin of the coordinate system. Theopposite would occur if it were beingenlarged.

y y

x x


22/120

Transformations 3 2

The cylinder appears to be largefrom the viewpoint of the camera

Skew -30 0 1 0 1 0 0

ie. lean the yaxis 30 toward the xaxis

Rotating and scaling an object that is NOT positioned at the origin of thecoordinate system can give rise to unexpected resu lts. Scaling, for examp le,

has the effect of mov ing the su rface of an object tow ard or aw ay from th e

origin d epend ing on w hether the object is being red uced or enlarged. If the

space into w hich an object is to be placed is translated, rotated an d/ or scaled,

it norm ally m akes more sense to ap ply the translation AFTER the rotation and

scaling as shown in the lower d rawing.

The following illustrations are based on Getting Started example 4, they

are intended to show how sw apping the order in w hich Scale an d Translate

are app lied results in the cylind er being p laced in tw o entirely d ifferent

locations in space. In each example the camera is located a t the origin of the

coordinate system.

Applying Transformations

Scale 0.3 0.3 0.3 #3rdTranslate 0 0 5 #2ndRotate -120 1 0 0 #1stCylinder 1 -1 1 360

1st 2nd

3rd

y

z

Translate 0 0 5 #3rdScale 0.3 0.3 0.3 #2ndRotate -120 1 0 0 #1stCylinder 1 -1 1 360

1st

2nd3rd

y

z

Skewing

Here the cylinder appears to bemuch smaller

y

x

y

x


23/120

Transformations 3 3

Applying Sequencies of Transformations

Display "3tubes" "framebuffer" "rgb"


WorldBeginTranslate 0 0 7Rotate -110 1 0 0Rotate 25 0 0 1

#insert the upright red cylinderScale .5 1 1Color 1 0 0Cylinder 0.5 -2 2 360

#insert the green second cylinderRotate 90 1 0 0

Scale .5 1 1Color 0 1 0Cylinder 0.5 -2 2 360

#insert the blue third cylinderRotate 90 0 1 0Scale .5 1 1Color 0 0 1Cylinder 0.5 -2 2 360

WorldEnd

RIB

The purp ose of this section is to gain a better und erstanding of the cumm u-lative effect of applying repeated sets of transformations to a modelling space.

The RIB file show n above illustrates one of the un expected resu lts that can

occur with transformations. In p articular n otice how the repeated scaling in

the x direction (shown in bold in th e RIB script) not only comp ound s the

overall squashing of each tu be but also redu ces the length of the blue tube.

This is quite an unexpected result. After all why should the length of th e last

tube be effected alon g its zaxis by scalings that h ave only been ap plied to th e

xaxes? The illustrations on the next two pages provide a step-by-step

explanation.


24/120

Transformations 3 4

x

The camera coordinate system may be considered to lie on the su rface of the

computer screen as if it were a large view-finder of a virtual camera . At the

beginning of the RIB script the axes shown below (a) mar k wh at is called thecurrent coordinate system.

The following RIB statements transform w hat m ay be thou ght of as a copy of

the current coordinate system before the first object is inserted into the w orld .

_3 Translate 0 0 7 _2 Rotate -110 1 0 0 _1 Scale 0.5 1 1

Cylinder 0.5 -2 2 360

As soon as th e cylinder is d eclared, the transformed copy of the coordinate

system (b) becomes th e current coordinate system. Subsequent transformations

will take p lace with reference to the new axes (c).

y

x

y

z

z

x

y

Notice how the xaxis has beenscaled by 50%.

Camera space

z y

z

x

(a)

(b)

(c)

_3

_1

_2


25/120

z

x

y

Transformations 3 5

The second set of RIB statements shown below transform w hat again m ay be

though t of as a copy of the current coordinate system.

Rotate 90 1 0 0Scale .5 1 1Cylinder 0.5 -2 2 360

Once again as soon as the second cylind er is declared, the copy of the

coordinate system (d) becomes the current coordinate system (e). Notice the

xaxis has received a fur ther red uction in scale.

z

y

x

The last set of RIB statements ap ply the final tran sformations prior to inser ting

the th ird cylind er.

Rotate 90 0 1 0Scale .5 1 1Cylinder 0.5 -2 2 360

Again th e xaxis of the m odelling space receives another redu ction in scale.

How ever, because of the rotation of the coordinate system around the yaxis,

the final cylinder is inserted p arallel to a coord inate axis that has been

considerably squashed hence the dram atic redu ction in th e length of the

cylinder.

z

y

x

z

y

x

The hollow arrows indicate thetransformed coordinates before

they become the current system.

(d)

y

z

(e)


26/120

The pu rp ose of this section is to introd uce the basics of rend ering objects

ph oto-realistically. It is intended to be a p ractical introd uction an d as su ch it

does not ad dress the theory of rendering.

To prod uce a realistic image a rend erer mu st be provided with information

about th e nature of the light sources and th e material and geom etric attributes

of the objects in a synthetic scene. In add ition, the rend erer also requires

information about th e virtual camera that is being u sed to view the scene. For

the sake of simplicity the examples in th is section use a camera w ith a

standard focal length lens, that provid es full coloured low resolution im ages,

eg.

Projection "perspective" "fov" 40Display "untitled" "framebuffer" "rgb"

Format 300 200 1

A simp le object from th e Rend erMan librar y of quad ric surfaces, eg.

Sphere 1 -1 1 360

is used through out this section so that any und ue complications that might

arise as a result of modelling a complex object(s) can be av oided .

If compu ter graph ics can be likened to painting by n um bers, then the

process of rendering a 3D scene can be th ought of as painting with light.

Therefore, a photo-realistic system such as RenderMan, whose purpose is to

create imagery th at is indistingu ishable from real ph otograph y, is useful only

to the extent that it enables users to make su btle changes to th e way light

interacts with each part of a scene.

At each stage in the rendering process the w ay that light is changed as a result

of these in teractions is called shading, and RenderMan p rovides a mechanism

for controlling th e outcome of each interaction th rough the u se of what it

describes as shaders. For examp le, the interaction of light with the su rface of an

object is controlled by a surface shader; while the cha racteristics of the light

prod uced by a sou rce of illum ination are governed by a light source shader. In

add ition to th ese two types of shad ers this section also introdu ces a shad erthat controls the way light reacts with an object whose surface has been

displaced to form sm all bum ps and pits a displacement shader.

Although Rend erMan p rovides seven d ifferent typ es of shaders, this section

only introdu ces three of them. The following RIB scripts have been carefully

selected to act as templates for your own experiments in the creative use of

Surface, Lightsource an d Displacement shaders.

Shading the basics

Shading the basics 4 1


27/120

Using lights and materials

The followin g scene consists of a coloured plastic sph ere that is illum inated

by two lights.

Projection "perspective" "fov" 40Display "untitled" "framebuffer" "rgb"Format 300 200 1

Translate 0 0 5Rotate -120 1 0 0Rotate 25 0 0 1

WorldBeginLightSource "ambientlight" 1 "intensity" 0.1LightSource "distantlight" 2 "intensity" 1.5 "from" [0 0 4] "to" [0 0 0]

Color 1 0 0Surface "plastic"Sphere 1 -1 1 360

WorldEnd

RIB

Of the two light sources used to illuminate th e scene the first provides a very

small amoun t of backgroun d, or am bient lighting. The second light is

positioned directly above the sph ere and is pointing toward the origin of the

coordinate system an d hence toward the centre of the sphere. Both "ambient-

light" and "distantlight" are provid ed by Rend erMan as p art of a basic library

of four typ es of light sou rces, the oth ers are "pointlight" and "spotlight".

The surface of the sphere has been a ssigned th e reflective proper ties of plastic

eg.

Surface "plastic"

Sphere 1 -1 1 360

(Using a shiny m aterial makes it somewh at easier to see how the high-lights

respond to any changes that are mad e to the p osition of the light sou rces.)

Rend erMan su pp lies a small nu mber of materials eg. plastic, wood , granite,

carpet etc., that can be u sed w ith the Surface statement. Each material has its

own way of being tun ed to the ind ividual requirements of a scene. For

example, the sph ere could have been given the p roperties of a very m atte and

non-reflective rough ened plastic with the following statement,

Surface "plastic" "Ks" 0.1 "roughness" 0.5

The surface shader "plastic" can use five parameters (the default valu es for

these are given in p arenth eses), namely,

"Ka" respon se to ambient light (1.0),

"Kd" d iffuse reflections (0.5),

"Ks" specular reflections (0.5),

"roughness" graininess of the surface (0.1), and finally,"specularcolor" the colour of the h igh-lights ([1 1 1]).



28/120

Applying an image to an object

The scene used in this examp le is very similiar to the first except th at a sur face

shader called "texmap" is used to wrap a 2D image around the sph ere a

technique know n as texture mapping,

Surface "texmap" "texname" ["your picture.tx"] "maptype" 2

The term texture in the context of 3D computer graphics is a little misleading

because it only refers to variations of the colour of a surface, it does not im ply

anything abou t its structure or rou ghness. Since there are a num ber of ways an

image can be wrap ped arou nd an object, "texmap" mu st be told to use

"maptype" 2 ie. sph erical mapp ing. Before "texmap" can apply an image to a

surface, the image mu st first be used to generate an interm ediate texture file.

The following statement does the necessay conversion,

MakeTexture "your picture.tiff" "your picture.tx" "periodic" "periodic""gaussian" 2 2

The main thing to note is that the image to be u sed as the source for the

texture file, which in th is instance is called "your picture.tiff", is located in the

same folder as the RIB file itself otherw ise the rend erer has n o way ofknowing where to find the appropriate file. Generally, picture files are either

created or modified using PhotoShop. It is essential they are stored as RGB

files rather th an, say, gray scale images. Once an image has been u sed to create

a texture file the MakeTexture statement can be commented-out.

The purp ose of the last part of the MakeTexture statement ie. "periodic"

"periodic" "gaussian" 2 2, is to allow the texture to be rep eatedly tiled over the

sphere should th at be necessary and to ensure the resulting texture map has a

smooth, or an ti-aliased, app earance. Like man y of the areas touched u pon by

this section, discussions about th e finer d etails of texture m app ing are d ealt

with elsewhere.


Display "untitled" "framebuffer" "rgb"Format 300 200 1


MakeTexture "your picture.tiff" "your picture.tx" "periodic" "periodic" "gaussian" 2 2

WorldBeginLightSource "ambientlight" 1 "intensity" 0.1LightSource "distantlight" 2 "intensity" 1.5 "from" [0 0 4] "to" [0 0 0]Surface "texmap" "texname" ["your picture.tx"] "maptype" 2

Sphere 1 -1 1 360WorldEnd

RIB



29/120

Preparing an image for texture mapping

Scan and/ or mod ify a grap hic using PhotoShop, save it as either a TIFF or a

PhotoShop 2.5 file. Even if it is a mon ochrome image be sure to m anip ulate it

in RGB mod e within Ph otoShop.

Redu ce the graph ic to a square format by choosing Im age Size from th e

Image m enu item. The Image Size dialog box will allow the graph ic to be

resized to a squ are asp ect ratio, say 800 x 800 pixels.

Select All using the Select men u item then rotate the graph ic 180 degrees

with the Rotate command und er the Image menu.

Use Save As to store the gr aph ic as a TIFF file in the sam e folder as th e

RIB file that w ill use it for texture map ping.

DO NOT MODIFY THE ORIGINAL GRAPH IC ALWAYS WORK ON ACOPY.

Although the Macintosh operating system does not require file extensions,

make sure the image file is named with a .tiff extension eg.me.tiff. Nam ing

image files and RIB scripts with .tiff and .rib extensions makes them very

easy to ident ify on the desktop. It is advisable to comp ress the file using LZW

compression.

Within the RIB script convert the graph ics file to a texture file with th e

statement,

MakeTexture "me.tiff" "me.tx" "periodic" "periodic" "gaussian" 2 2

Once the RIB script has been used successfully the MakeTexture statement can

be comm ented-out ie.

#MakeTexture "me.tiff" "me.tx" "periodic" "periodic" "gaussian" 2 2

This will ensure that su bsequent rend erings will be comp leted as quickly as

possible. Of cour se if the original grap hics file is altered th en a new textu re file

mu st be produ ced, in w hich case the comm ent (ie. #) must be removed .

Use the surface shader texmap to wr ap th e texture file around the sphere.

Surface "texmap" "texname" ["me.tx"] "maptype" 2Sphere 1 -1 1 360

The texmap surface shader allows the same p arameters as the p lastic

shad er (ie. Ka, Kd , Ks, roughness and specularcolor) to be used to cont rol the

way light reflects from the surface of the texture map.

Step 1

Step 2

Step 3

Step 4

Step 5

Step 6



30/120

Using an image to displace a surface

In this example an image is u sed to u pset or d isplace the surface of a sphere

a technique know n as displacement mapping. Similiar to the su rface shad er

texmap, the emboss displacement shader requires an image file to be

converted to an intermediate texture file ie.

MakeTexture "your picture.tiff" "your picture.tx" "periodic" "periodic""gaussian" 2 2

Again you m ust ensure th e image to be used as the source for the displace-

men t m ap is located in th e same folder as the RIB file itself. A tiff file to be

used as a source image for displacement map ping can be saved in either

PhotoShops gray-scale or rgb mode. However, gray scale images give a more

pronou nced em bossing effect. Once an im age has been used to create a texture

file the MakeTexture statement can be commented-out eg.

#MakeTexture "your picture.tiff" "your picture.tx" "periodic" "periodic""gaussian" 2 2

Unlike a techniqu e called bu mp map ping, displacement m app ing really

does m ake changes to the geometry of the su rface to w hich it is app lied. It is,

therefore, superior to bum p m app ing which only makes a surface appear to be

bumpy. The emboss displacement shad er mu st be given the nam e of a

texture file that it w ill use for em bossing ie.

Displacement "emboss" ["your picture.tx"] "Km" 0.03

It should also be given a reasonable value for the param eter Km w hich sets

the magn itude of the embossing. The emboss shader respond s to the gray

values in the texture file lighter p arts of the image are pressed deeper in to

the su rface. The default va lue for Km is 0.03.


Display "untitled" "framebuffer" "rgb"Format 300 200 1


MakeTexture "your picture.tiff" "your picture.tx" "periodic" "periodic" "gaussian" 2 2

WorldBeginLightSource "ambientlight" 1 "intensity" 0.1LightSource "distantlight" 2 "intensity" 1.5 "from" [0 0 4] "to" [0 0 0]Displacement "emboss" "texname" ["your picture.tx"] "Km" 0.03

Sphere 1 -1 1 360WorldEnd

RIB



31/120

Avoiding rendering errors and improving performance

The renderer u ses each objects boun ding box to qu ickly determine wh ere their

surfaces are located in a scene. In th is way it avoids trying to render emp ty

space. How ever, wh en the su rface of an object is shifted a s a result ofdisplacement m app ing, PRMAN m ay make severe rendering errors. The

renderer literally ignores those parts of the object that have been displaced

outside their bound ing box. Rend erMan provid es a mechanism by w hich the

renderer can be warned about such d isplacements eg.

Attribute "bound" "displacement" [0.2]Displacement "emboss" ["your picture.tx"] "Km" 0.03

Shere 1 -1 1 360

As long as it app ears before the nam e of the object, Attribute can be inserted

before or after the Displacement statement. Unfortunately, even thou gh the

displacement m agnitud e is set with the Km factor their is no way of knowing

the exact Attribute value to u se. In th e example shown above it is set to 0.2 but

often 0.1 is enough to p revent rend ering errors from occuring.

Rend ering operations that involve texture files require more m emory th an

those that d ont. To ensure it can operate on compu ters with m odest amou nts

of memory, Rend erMan only sets aside a small amou nt of mem ory for

working w ith textures. To improve p erformance an option m ay be set to

inform the rend erer to work w ith larger chu nks of texture ie.

Option "limits" "texturememory" [4096]

The valu e 4096 specifies the nu mber of Kbytes (4 MB) to set asid e for

mem ory to be u sed to store information read from a texture file. If you w ish to

use th is option p lace it at the beginn ing of the script options effect the whole

scene and mu st be set before the camera and world are described.


Bounding boxes

Memory and speed


32/120

This section explores the ways that objects are defined in a virtual world.

Because our 3D w orlds are d escribed by han d w ritten RIB files they will be

relatively simp le. However, this is not a disad van tage because it will focusattention on imp arting as mu ch visual interest through th e use of careful

shading techniques and sensitive lighting, rather than gratutious comp lexity

obtained all too easily by the use of an interactive mod elling system . Before

embarking u pon the intricacies of lighting and shading som e competence

mu st be gained with m odelling. This section is designed to provide you with

these skills.

Shap ing Up takes an in-dep th look at two typ es of surfaces comm only used

to construct virtual models, namely, quadrics an d polygons. Sophisticated

modellers also use surfaces based on curves called splines. If you have used

an illustration program such as Adobe Illustrator or Aldus FreeHand youwou ld h ave emp loyed 2D splines to create curves. However, 3D splines are an

advan ced top ic of stud y and will not be addressed in this course.

Particular u se will be made of the library of shapes, or primitives as they are

sometimes called, that are built into RenderMan. These pre-defined surfaces

are based u pon mathem atical expressions called qu adratic equations, hence

their general name ofquadric surfaces. There are seven su rfaces in th e library

and they are illustrated on th e next two pages. Each quad ric has its own set of

parameters that allow its form to be accurately specified. The meaning of these

param eters and examples of their u se are given. In ad dition to being d escribed

by an equation they are also surfaces of revolution. That is, they are formed byspinning a line or curv e around a central axis. Most mod elling program s offer

these primitives because it is easy to assemble them into composite models.

Unlike man y renderers Rend erMan does not ap proximate quad rics in any

way and so renders them with smooth silhouettes.

The other typ e of sur face that w ill be used is a polygon - a flat shape enclosed

by straight edges. Traditionally, polygons have been very important in 3D

compu ter graph ics because of the ease with w hich th ey can be

internally represented by mod ellers and rend erers,

assembled into a skin or mesh that approximates a desired form, and rendered in a variety of ways to give the illusion of smoothness.

The straight ed ges of a polygon are d efined by a sequen ce of 3D vertices each

of which is specified by three n um bers its x, y and z coordina tes. Since even

simple polygon m eshes can consist of dozens of polygons each consisting of

at least three vertices (ie. triangles), it will only be feasible for us to describe

very simple surfaces.

Shaping Up library objects and polygons

Overview

Shaping Up 5 1

Quadrics

Polygons


33/120

z

y

x

theta max

zmax

zmin

radius

RenderMans Library of Quadric Surfaces

z

y

x

z

y

x

theta max

height

radius

z

y

x

theta max

zmax

zmin

radius

height

radius

theta max

Disk 0.5 1.0 300 (example)

Cone 1.75 1.0 270

Sphere 1.0 -0.5 0.8 270

Cylinder 1.0 -0.5 1.0 300

Cone height radius thetamax

Cylinder radius zmin zmax thetamax

Sphere radius zmin zmax thetamax

Disk height radius thetamax (syntax)

Shaping Up 5 2


34/120

Shaping Up 5 3

y

zmax

zmin

z

y

x

theta maxpoint1

point2

z

y

x

theta max

major rad

min rad

phi min

phi max

x

z

theta max

RenderMans Library of Quadric Surfaces - continued

Torus 1.0 0.3 90 320 300

Paraboloid 1.0 0.15 1.2 300

Hyperboloid -0.3 1.0 -1.0 0.7 0.7 1.0 300

Torus major rad min rad phimin phimax thetamax

Paraboloid radius zmin zmax thetamax

Hyperboloid point1 point2 thetamax

radius


35/120

RIB

Shaping Up 5 4

Example 1 - dont forget the inside!

#better goblet.RIB#adding an inside surfaceDisplay "goblet" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1

WorldBeginLightSource "pointlight" 1 "intensity" 40 "from" [4 2 4]Translate 0 0 5Rotate -120 1 0 0

Surface "plastic"Color 1.0 0.9 0.3 #goldCylinder 1 0 1.5 360 #containerDisk 0 1 360 #base of the containerCylinder 0.25 -1.5 0 360 #stem

Disk -1.5 1 360 #base of the goblet

Translate 0 0 1.5 #move the origin to the top of the gobletSphere 1 -1 0 360 #hemi-spherical inside surface

WorldEnd

This example introd uces the first of the library shapes a sph ere. It also uses

two new RIB statements, LightSource an d Surface. 3D compu ter graph ics

has d eveloped a rich set of lighting and surface texturing techniques that can

dramatically alter the appearance of an object. Although the concepts are

dealt with in detail in later sections, light sources and material attributes canstill be used effectively, even withou t elaborate explan ations, to add realism

to a model.

With th e exception of those lines marked in italics, this file is the sam e as the

final example of the prev ious section. At the end of the scene d escription the

origin is moved to the top of the goblet and th e lower half of a sphere is

placed within th e container by the RIB command ,

Sphere radius zmin zmax thetamax

A (poin t) light sou rce is oriented to high-light the curved sur faces of the

goblet. The harshness of the lighting can be reduced by inserting this line,

LightSource "ambientlight" 2 "intensity" 0.2

imm ediately after the first light source statement. The RIB comm and Surface,

followed by the n ame of a m aterial in th e Rend erMan library acts much like

Color intha t all subsequen t objects acqu ire the chosen characteristics.

Although m ore will be said about m aterials and surface textures, you may

like to experiment by su bstituting the par ameter p lastic for any on e of

those shown in th e list given opposite. Later you will be shown how to

control the char acteristics of each ma terial.

carpetcloth

cmarble

constantfinemetal

Mattemetal

paintedplasticplastic

rmarblersmetal

shinymetalspatter

stonewood


36/120

Shaping Up 5 5

Placing objects in the world

Cylinder 1 0 1.5 360

RIB

The actions of some of the RIB statemen ts in the first examp le are illustrated

below. In each case the position of the xy p lane is indicated by a grid . The

surface being created is shown in the heavier line weight and the par ameter(s)responsible for positioning th e surface in th e z d irection are show n in bold.

Disk 0 1 360

Cylinder 0.25 -1.5 0 360

Disk -1.5 1 360

Translate 0 0 1.5Sphere 1 -1 0 360


37/120

Shaping Up 5 6

Example 2 - adding a rim and moving the camera

#goblet with rim.RIB#adding a rimDisplay "goblet" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1


WorldBeginLightSource "pointlight" 1 "intensity" 50 "from" [4 2 4]LightSource "ambientlight" 2 "intensity" 0.2

Surface "plastic"Color 1.0 0.8 0.3 #goldCylinder 1 0 1.5 360 #container

Disk 0 1 360 #base of the containerCylinder 0.25 -1.5 0 360 #stemDisk -1.5 1 360 #base of the goblet

Translate 0 0 1.5 #move the origin to the top of the gobletCylinder 0.9 -1.4 0 360Disk -1.4 0.9 360Torus 0.95 0.05 0 180 360

WorldEnd

Although this examp le demonstrates the u se of a torus, its main feature is the

way the transformations,


that were previously used to rotate and m ove the goblet within the world

space are now effecting the whole wor ld. Remember, all RIB statemen ts pr ior

to WorldBegin refer to the w ay the w orld is oriented with respect to the

camera. Because it makes more sense to change the cameraworld

relationship, as shown by the illustration on the next page, it will no longer be

necessary to rotate and move ind ividual objects to obtain a better view of

them.

After a cylindrical liner and a flat base have been add ed to the inside of the

goblet a round ed rim is created w ith the Torus statement,

Torus major rad min rad phimin phimax thetamax

Try to ad d an other d isk to the base of the goblet and provide it w ith either a

RIB


38/120

x

y

z

yx

z

Shaping Up 5 7

Display "goblet" "framebuffer" "rgb"


RIB

display the graphic in a window titled

"goblet", 200 by 150 pixels in size,use a camera with a 40 degree field ofvision and include rgb colour data

initially the origins of the cameraand the world coincide

x

y

z

y

x

z

rotate the world -120 degreesaround the x axis

move the world 5 units along the zaxis of the camera

since transformations occur inreverse order the rotation is followedby the translation

x

y

z

y x

z


"freeze" the camera - now only usethe world coordinates

WorldBegin(assemble the goblet)

x

y

z yx

z

WorldEnd scene description complete

Positioning the w orld relative to the camera


39/120

#coffee mug.RIB#modifying the gobletDisplay "mug" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1

Translate 0 -0.5 5Rotate -120 1 0 0Rotate 45 0 0 1

WorldBeginLightSource "pointlight" 1 "intensity" 50 "from" [4 4 4]LightSource "ambientlight" 2 "intensity" 0.25

Surface "plastic"Color 0 0 1 #fully saturated blue

Cylinder 1 0 1.5 360 #mugDisk 0 1 360 #base of the mugTranslate 0 0 1.5 #move the origin to the topCylinder 0.9 -1.4 0 360 #lining of the mugDisk -1.4 0.9 360 # bottom of the mugTorus 0.95 0.05 0 180 360 #mug rim

Translate 0 1 -0.75 #move the origin to the back, and lower it half way down the mugRotate 90 0 1 0 #rotate the origin so that the handle will be verticalTorus 0.6 0.1 0 360 180 #create a handle

WorldEnd

In this example some minor alterations to the scene have changed the goblet

into a coffee mug. The statements relating to the stem and base have been

removed an d those shown in bold have been add ed or altered. How ever, the

most imp ortant p oint to notice about this file is the way th e world is rotated 45

degrees clockwise abou t the z axis before it is tipp ed back 120 degrees. In all

the p revious examp les the camera w as vertically aligned w ith the y axis of the

world . If you place a comment in front of the cameras second rotation you

will immed iately see the effect it has on th e view. In add ition, the m ug has

been centred by moving th e world 0.5 units dow n th e y axis of the camera.

Introduce a Scale statement to widen the hand le as shown. The mu g does not

look tall enough increase its height to 1.9 un its.

RIB

Example 3 - anyone for coffee?

Shaping Up 5 8


40/120

Cylinder 1 0 1.5 360 Disk 0 1 360

yy

y

y

y

y

x x

x

x

x

x

Translate 0 0 1.5Cylinder 0.9 -1.4 0 360

Disk -1.4 0.9 360

Torus 0.95 0.05 0 180 360 Translate 0 1 -0.75Rotate 90 0 1 0Torus 0.6 0.1 0 360 360

The actions of the RIB statements u sed in th e construction of the coffee mug ,

examp le 3, are illustrated below. Unless otherwise ind icated , the z axis is

pointing u p. The surface being created is shown in the h eavier line weight andthe parameter(s) responsible for positioning the suface in the z direction are

shown in bold.

In the last diagram th e hand le has been w idened by app lying a scaling factor

to the x coordinate. It is left as an exercise for you to d etermine wh ere in the

script th e Scale 2 1 1 command should be inserted.

z

y

x

z

Shaping Up 5 9

3

5

7

1

4

2

6

Scale 2 1 1

Question: why is the scaling being applied

to the x axis when in this diagram it app ears

as if the z axis requires stretching?


41/120

Shaping Up 5 10

#saucer.RIB#some tricky scalingDisplay "saucer" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1


WorldBeginLightSource "pointlight" 1 "intensity" 45 "from" [2 -3 4]LightSource "ambientlight" 2 "intensity" 0.15

Surface "plastic"Color 0.5 0.5 1 #pale blue

Translate 0 0 0.5Scale 4.4 4.4 1

Sphere 0.5 -0.5 0 360Sphere 0.4 -0.4 0 360Torus 0.45 0.05 0 360 360

WorldEnd

The pu rpose of this example is to show the effect of using scaling an d trans-

lation, as well as the importan ce of app lying th ese transformations in the

correct order. Ignoring the tran sformations for a mom ent, the combination of

the two sp heres and th e torus simply p rodu ces a hemi-spherical cup , 0.1

un its in thickness, with a round ed rim.

The effect of the Scale statement is to stretch the cup into a saucer-like

object, refer to the illustr ation on th e next page. The translation is opt ional in

that it d oes not change th e form of the saucer, only its position. How ever, it

makes sense to lift the saucer by an amou nt equ al to its radius so that it sits

on the xy plane, hence the translation of 0.5 un its in the z direction.

The widened rim of the saucer is du e to the thickness of the basic cup being

exaggerated by the scaling factorlike stretching a sheet of rubber. By ad just-

ing the d iameter of the inner sp here, and m aking the necessary changes to the

param eters of the torus, a wide var iety of rims can be created.

The basic cup can also be stretched v ertically into an object reminiscent of

an egg cupsee the next page. To create this object, x and y have been scaled

by 1, therefore they remain un changed, wh ile the height in th e z d irection has

been increased by 200%. To comp ensate for the scaling, the tran slation has

been increased from 0.5 to 1 unit ie. 2 x 0.5 = 1.

As an exercise, create an egg by scaling a sph ere, assign it an ap prop riate

colour an d p osition it in the egg cup .

RIB

Example 4 - the universal saucer


42/120



no translation

Shaping Up 5 11





The effects of scaling and translation

no scaling or translation

both scaling and translation ap plied

Translate 0 0 1Scale 1 1 2


y

y

x

x

y

x

y

x

now its an egg cup !

For reference the or iginal xy p lane BEFORE the transform ations w ere app lied

are shown in each example.


43/120

Shaping Up 5 12

#eggcup with base.RIB#copying and pasting with instancingDisplay "eggcup" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1

ObjectBegin 1Sphere 0.5 -0.5 0 360Sphere 0.4 -0.4 0 360Torus 0.45 0.05 0 360 360

ObjectEnd

Translate 0 -.5 5Rotate -120 1 0 0Rotate 60 0 0 1

WorldBeginLightSource "pointlight" 1 "intensity" 25 "from" [2 -3 4]LightSource "ambientlight" 2 "intensity" 0.25

#Egg cup topColor .55 .17 .11 #dark brownSurface "wood"Translate 0 0 1.0Scale 1 1 2ObjectInstance 1

#Egg cup baseTranslate 0 0 -0.5Scale 1 1 0.25

Rotate 180 1 0 0ObjectInstance 1

WorldEnd

RIB

Example 5 virtually greenreusable geometry

The previous examp le illustrated an imp ortant p oint about 3D mod els; by

making a few m inor changes, to scaling for examp le, their geometry can form

the basis of a variety of secondar y models. A similiar principle can be app lied

within a single scene description. This example shows how several surfaces

can be collected together into a sing le retained object, and conveniently reused,

or instanced, many times. In th e context of a dr awing program this is like mak-

ing a group, then copying and pasting it repeatedly w ithin an illustration.

The intention to make an ob ject from a collection of surfaces is indicated to the

renderer by ObjectBegin/ObjectEnd. The n um ber following ObjectBegin

indentifies, or tags the collection for later use by a statem ent tha t places the

object in the wor ld, the nu mber itself has no oth er significance,

ObjectInstance tag

Unfortunately, transformations cannot be used between ObjectBegin an d

ObjectEnd. Instancing allows the renderer to work more efficiently and also

helps to avoid writing tediously long RIB files. Pay particular attention to the

next two p ages as they explain th e transformations used in this scene.


44/120

The following line d rawings sh ow the effect of app lying th e transformations

used in example 5. At each stage, the coordinate system is represented by a

one un it grid, subdivided into quarters. To fully un derstand the action ofeach group of transformations, remember they are app lied,

in reverse sequence, and

with reference to the current coordinate systemshown as heavier lines.

The new coordina te system only becomes current when an object is created.

untransformed Scale 1 1 2 Translate 0 0 1 ObjectInstance 1

Shaping Up 5 13

Rotate 180 1 0 0 Scale 1 1 0.25 Translate 0 0 -0.5 ObjectInstance 1

Translate 0 0 1Scale 1 1 2ObjectInstance 1

RIB (fra gmen t)

Translate 0 0 -0.5Scale 1 1 0.25Rotate 180 1 0 0ObjectInstance 1

RIB (fra gmen t)

Visualising example 5


45/120

In the last example, the base of the egg cup w as positioned an d p laced in th e

scene by a scaling followed by a tran slation.

The same effect can also be achieved by p lacing th e translation before the

scaling. H owever, simply reversing the two statements w ill not w ork. As the

draw ing below show s, the translation m ust be altered. In general it is better to

perform a scaling BEFORE a translation, as shown on the previous page.

Shaping Up 5 14

Rotate 180 1 0 0 Scale 1 1 0.25Translate 0 0 -2 ObjectInstance 1

-1

-2

Visualising example 5 continued

RIB (fra gmen t)

#Egg cup baseScale 1 1 0.25Translate 0 0 -2Rotate 180 1 0 0

ObjectInstance 1


46/120


47/120

happ ened in example 1, or reset as shown. Three surface shaders are used in

this example to give the effect of a carpet, wood and spatter. Their full

specifications, as documented by PIXAR, appear on the next three pages.Although the d etails of surface shading will be dealt with in an other section,

the d escriptions of each of the m aterials used here should give you enou gh

information to und ertake your own experiments. Most of the param eters use

values that ran ge from 0 to 1; the exceptions, at least for the m aterials used in

this example, are

wood "grain"

carpet "nap"

spatt er "sizes"

Experiment w ith some or all the param eters in ord er to app reciate the control

that each provides over the ap pearance of a surface. In the absence of pre-

compu ted im ages this is a trial and error p rocess. Three techniqu es can be

used to speed up rendering

use a higher ShadingRate, say 20, and opt ionally use

ShadingInterpolation smooth, to red uce the blotchiness of the image,

comment-out any surfaces that are not currently being adjusted.

ShadingRate is like a quality control ad juster ; low valu es of around 1 or 2

give excellent results w hile higher values like 20 or m ore provide rough and

ready snap shots. A high shad ing rate simply tells the renderer not to cal-

culate the colour value for every pixel but to sample the pixels at whateverrate has been set. The closest comparison to real world photography is

choosing a high sp eed film with a coarse grain emu lsion. Unfortunately, high

shading rates generate very pixelated images, see opposite. To reduce these

artefacts,

ShadingInterpolation "smooth"

can be used to tell the renderer to average-out, or interp olate, the p ixels

between the samples, otherwise it simply uses a constant colour. The state-

ment can be inserted immed iately after ShadingRate. Of course you may

wish to take adv antage of these image d efects to achieve a par ticularillustrative effect, in w hich case resetting ShadingInterpolation is optional. By

defau lt its set to "constant", hence the b locks of flat colour.

Bearing in m ind a PAL resolution video image constists of 442,368 pixels, the

careful u se of these statements can have a very significant effect on the speed

of rendering.

Shaping Up 5 16


48/120

Shaping Up 5 17

"wood" "Ka" "Ks" "Kd" "roughness" "specularcolor" "grain" "swirl"

"swirlfreq" "c0" "c1" "darkcolor"

"Ka" 1 "Ks" 0.4 "Kd" 0.6 "roughness" 0.2"grain" 5 "swirl" 0.25 "swirlfreq" 1"specularcolor" [1 1 1]"darkcolor" [dependent on the surface colour]

"c0" [0 0 0] "c1" [0 0 1]

This shader creates a realistic-looking wood . The frequen cy of the w ood grain

can be changed with the grain param eter. The relative amou nt or am plitude of

the tur buent sw irl in the grain is controlled by the swirl param ater, and swir-

freq controls the frequency of this turbulence. Low values of swirl produce

more u niform looking w ood, wh ile low values of swirlfreq m ake the wood

app ear to be m ore knotty. Obviously these two param eters interact to a largeextent. You sh ould be careful not to set swirl too high or sw irlfreq too low or

the wood will become a jumbled m ess.

The wood is simu lated by creating a grain that is essentially composed of

differently coloured concentric cylinders around a central axis defined by the

two p oints c0 an d c1. This axis is the z axis by d efault. Note th at the or ientat-

ion of this axis can be var ied either by changing these two param eters or by

doing some tran sformations between th e call to the shader an d th e definition

of the geometry. Either one of these ap proaches may make more intu itive sense

in different applications.

The colour of the wood will norm ally consist of band s of d ifferent intensities of

the su rface colour. This is the most generally useful way of invoking the

shader. However, for special appearances this can be changed by changing the

darkcolor param eter, which controls the colour of the dark grain of the w ood.

The different intensity levels are actually levels of mixing betw een this colour

and the sur face colour, so setting the su rface colour to red and darkcolor to

wh ite will produce red wood with w hite grain and various shades in between.

The parameters Ka, Ks an d Kd have the u sual meanings of ambient, specular

and d iffuse reflective in tensities, respectively. roughness an d specularcolor

control the sharpness and colour of the specular highlight.

This shad er can have p roblems w ith aliasing.

Name

Defaults

Description

Bugs

Surface Shaders


49/120

Shaping Up 5 18

"carpet" "Ka" "Kd" "scuff" "nap"

"Ka" 0.1 "Kd" 0.6 "scuff" 1 "nap" 5 "swirl" 1

This shader p rodu ces a carpeted su rface, comp lete with scuff-marks. The scuff

parameter controls the amount of scuff, or the relative frequency of intensity

variations. Higher values prod uce more frequent scuffing. nap describes the

shagginess of the carpet. Higher valu es make a m ore coarse-looking carpet.

The carpet shader makes a reasonable stab at anti-aliasing, so the actual grain

of the carpet fades away w ith distance.

There are no sp ecular reflections from rea l carp et (at least on a macroscopic

scale), so the only lighting parameters are Ka an d Kd, which have the usual

mean ings of ambient and d iffuse reflective intensities, respectively.

The way an ti-aliasing is perform ed can cause linear ar tifacts in some cases.

Name

Defaults

Description

Bugs

Surface Shaders continued


50/120

"spatter" "Ka" "Ks" "Kd" "roughness" "specularcolor" "basecolor"

"spattercolor" "specksize" "sizes"

"Ka" 1 "Ks" 0.7 "Kd" 0.5 "roughness" 0.2"specularcolor" [1 1 1] "basecolor" [0.1 0.1 0.5] "spattercolor" [1 1 1]"specksize" 0.01 "sizes" 5

This shad er m akes objects look like blue camp cookware with wh ite paint

spatt ers. Actually, both the blue basecolor and the wh ite spattercolor can be

changed if you d esire.

The parameter specksize controls the size of the paint specks as you w ould

expect. However, there are a ran ge of sizes of paint specks controlled by theparam eter sizes. Lower (integer) values p rodu ce smaller and more u niform

specks. Higher values p rodu ce some larger blotches and specks of man y

different sizes.

The parameters Ka, Ks an d Kd, have the u sual meanings of ambient, specular

and d iffuse reflective in tensities, respectively. roughness an d specularcolor

control the sharpness and colour of the specular highlight.

This shad er can have p roblems w ith aliasing.

Name

Defaults

Description

Bugs

Surface Shaders continued

Shaping Up 5 19


51/120

Shaping Up 5 20

RIB

Example 7a - making a compos ition the w rong way!

In this example the d escription of the saucer in example 4, show n in bold , has

been copied and pasted into the p art of the p revious RIB file that d escribed

the so-called table cloth. However, as the illustration on the next page shows,

something very strange has happen ed to the egg cup and egg.

#egg and cup.RIB#playing with materials

Display "eggncup.tiff" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1ShadingRate 5


ObjectEnd

Translate 0 -0.7 4Rotate -120 1 0 0

Rotate 60 0 0 1

WorldBeginLightSource "pointlight" 1 "intensity" 20 "from" [2 -3 4]LightSource "pointlight" 1 "intensity" 8 "from" [2 3 2]LightSource "ambientlight" 2 "intensity" 0.15

Surface "plastic"Color 0.5 0.5 1 #pale blueTranslate 0 0 0.5Scale 4.4 4.4 1

Sphere 0.5 -0.5 0 360Sphere 0.4 -0.4 0 360

Torus 0.45 0.05 0 360 360

#TopColor 0.55 0.17 0.11Surface "wood" "darkcolor" [0 0 0] "swirl" 0.25 "grain" 15 "swirlfreq" 1.5Translate 0 0 1.0Scale 1 1 2ObjectInstance 1

#EggSurface "spatter" "basecolor" [0.87 0.66 0.6] "sizes" 3 "spattercolor" [0.55 0.17 0.11]"Ks" 0.0 "Kd" 1Sphere 0.4 -0.4 0.4 360

#BaseColor 0.55 0.17 0.11Surface "wood" "darkcolor" [0 0 0] "swirl" 0.25 "grain" 15 "swirlfreq" 1.5Translate 0 0 -0.5Scale 1 1 0.25Rotate 180 1 0 0ObjectInstance 1

WorldEnd


52/120

Shaping Up 5 21

Because the saucer is created w ith a scaled coordinate system all the surfaces

defined after this transform ation are likewise effected , hence the Ostrich egg

effect shown on the left, rather than the d esired comp osition shown on theright!

Clearly objects in a scene need to have their individual coordinate systems, or

object space, and their surface attributes kept, in a sense, private from each

other. In a RIB file there are tw o ways in w hich this can be achieved . In th e

following example the two principle objects, the saucer and the egg cup

holding an egg, are blocked together between the statements AttributeBegin/

AttributeEnd; these instru ct Rend erMan to localise (keep pr ivate) the

geometry AN D the su rface attributes of each object. If only the geometry

needs to be kept p rivate, and there are good reasons why this is sometimes

necessary, then the TransformBegin/TransformEnd statements are used

instead.

To d raw an analogy, ifWorldBegin/WorldEnd define the beginning and end

of an entire theatr ical play, then the AttributeBegin/AttributeEnd behave like

markers that separate one scene from another. Surfaces and polygons fulfill

the role of actors with each, either separ ately or in collections, being

assigned costum es represented by the attribu tes ofColor an d Surface.

In the im proved RIB script on the following page, ObjectInstance has been

used to insert the sau cer instead of declaring three separate surfaces. None-

theless, the composition still displays a modelling errornotice how the egg

cup partially penetrates the saucer. How ever, because they are group ed

together with the AttributeBegin/AttributeEnd statements, the egg cup and

egg can be raised by a single transformation (show n in bold p rint on the n ext

pag e). Because of the curvatu re of the saucer a similiar error accurs if the egg

cup is moved toward the rim. But even so, it is sometimes acceptable to allow

objects to interp enetrate as long as th e error is not too not icable.

The two m ethods of group ing objects together using AttributeBegin/End an d

TransformBegin/End are summ arised on page 24.


53/120

Shaping Up 5 22

#egg cup and saucer.RIB

#combining objects the correct way!

Display "eggncup.tiff" "framebuffer" "rgb"Projection "perspective" "fov" 40Format 200 150 1ShadingRate 5


ObjectEnd

Translate 0 -0.7 4Rotate -120 1 0 0

Rotate 60 0 0 1

WorldBeginLightSource "pointlight" 1 "intensity" 20 "from" [2 -3 4]LightSource "pointlight" 1 "intensity" 8 "from" [2 3 2]LightSource "ambientlight" 2 "intensity" 0.15

AttributeBegin #SaucerSurface "plastic"Color .5 .5 1Translate 0 0 0.5Scale 4.4 4.4 1ObjectInstance 1

AttributeEnd

Translate 0 0 0.1 #raise the egg cup and egg

AttributeBegin #Egg cup and eggColor 0.55 0.17 0.11Surface "wood" "darkcolor" [0 0 0] "swirl" 0.25 "grain" 15 "swirlfreq" 1.5Translate 0 0 1.0Scale 1 1 2ObjectInstance 1

#EggSurface "spatter" "basecolor" [0.87 0.66 0.6] "sizes" 3 "spattercolor" [0.55 0.17 0.11]"Ks" 0.0 "Kd" 1Sphere 0.4 -0.4 0.4 360

#BaseColor 0.55 0.17 0.11Surface "wood" "darkcolor" [0 0 0] "swirl" 0.25 "grain" 15 "swirlfreq" 1.5Translate 0 0 -0.5Scale 1 1 0.25Rotate 180 1 0 0ObjectInstance 1

AttributeEndWorldEnd

RIB

Example 7b - making a composition the correct way


54/120

Shaping Up 5 23

WorldBegin

#lighting setup the same as before

Color 0.87 0.71 0.51Surface "carpet" "Kd" 0.8 "nap" 0.8 "scuff" 0.8TransformBegin #mug

Cylinder 1 0 1.5 360Disk 0 1 360Translate 0 0 1.5Cylinder 0.9 -1.4 0 360Disk -1.4 0.9 360Torus 0.95 0.05 0 180 360

#mug handleScale 2 1 1Translate 0 1 -0.75Rotate 90 0 1 0Torus 0.6 0.1 0 360 180

TransformEnd

Translate 0 0 -0.15 #lower the saucer

Color .65 .27 .21Surface "wood" "darkcolor" [0 0 0] "swirl" .25 "grain" 15 "swirlfreq" 1.5TransformBegin #Saucer

Translate 0 0 0.5Scale 4.4 4.4 1ObjectInstance 1

TransformEndWorldEnd

Example 7c - another way of grouping objects

RIB fra gmen t

This example is a combination of coffee mug.RIB an d saucer.RIB. It show s

how the surfaces and transformations that form an object can be grouped

together using TransformBegin/TransformEnd. Like the group s formed by

AttributeBegin/ AttributeEnd in the previous examp le, these new statements

keep the tran sformations for each object pr ivate, but th ey do so withou t

localizing colour and surface attributes. The image on th e left show s wh at

happ ens if the transformations previously applied to the h and le are not kept

pr ivate. The imag e on the righ t is the result of the RIB given above.


55/120

Independ ent objects each w ith

their own shape, position AND

shading attributes.

Shaping Up 5 24

Summary of methods relating to the grouping of objects

WorldBegin

AttributeBegin(shape, transformation and shading informationrelating to the egg and egg cup)

AttributeEnd

AttributeBegin(shape, transformation and shading informationrelating to the saucer)

AttributeEndWorldEnd

WorldBeginColor 1 1 0 #YellowTransformBegin

(only shape and transformationinformation relating to a saucer)

TransformEnd

TransformBegin(only shape and transformationinformation relating to a saucer)

TransformEnd


TransformEnd


TransformEnd


TransformEnd

WorldEnd

Independ ent objects each with

their own shape and position, BUT

all sharing a common shad ing

attribute wh ich in this example

has been set to the colour yellow.

AttributeBegin

AttributeEnd

TransformBeginTransformEnd


56/120

In this section the characteristics of a virtu al camera will be furth er refined to

include,

depth of field, and

motion blur.

In the previous chapters, RIB scripts have typically defined a camera with the

following statements,

Display "mypicture" "framebuffer" "rgba"Projection "perspective" "fov" 40Format 800 800 1

Such a camera, however, only approximates the behaviour of a real camera. In

pa rticular it acts like a pin-hole camera in th at it sharp ly focuses all parts of an

image. Real cameras are strictly limited in their ability to simultaneou sly focus

the images of objects placed at d ifferent distan ces in front of the lens this is a

ph ysical limitation that d epend s solely on the ratio of the d iameter of the

aperture of the cameras lens to its focal length. In photography this ratio is

called th e f-stop or f-number of a lens. This opt ical limitation is

aesthetically exploited by photograph ers wh en the foreground and / or the

background of a comp osition is d eliberately d efocused.

Real camer as are likewise un able to shar ply focus on objects in a scene if either

the camera is moving, particularly if it is hand -held, or if the scene, or par ts of

it, are in motion. To accomoda te wh at, in compu ter grap hics, is called m otionblur th e specification of a synthetic camera mu st includ e the idea of an image

being form ed ov er a finite period of time ie. as if using a shutter. The basic

virtual camera used so far has assum ed all motion is frozen by capturing an

instantaneous rep resentation of a synth etic scene.

If ph oto-realism is to be achieved in th e digital dom ain then similiar facilities

should be available to 3D comp uter gr aphics. By p roviding m echanisms to

introduce, what are in effect, digital versions of depth of field and motion

blur, RenderMan greatly im proves a designers ability to deliver ph oto-realistic

imagery. Interestingly, some real world photographic nuances are not available

in RenderMan, these includ e effects caused by,

lens flare,

optical defects such as astigmatism and chromatic aberration, and

the mu nd ane effects of dirty or scratched optics!

Because there is a slight d ifference in th e way field of vision (fov) is defined

in ph otography compared to Rend erMan, this section also provides some

examples of converting from on e system to th e other.

Overview

An Improved Camera 6 1

In this section the characteristics of a virtu al camera will be furth er refined to

include,

depth of field, and

motion blur.

In the previous chapters, RIB scripts have typically defined a camera with the

following statements,

Display "mypicture" "framebuffer" "rgba"Projection "perspective" "fov" 40Format 800 800 1

Such a camera, however, only approximates the behaviour of a real camera. In

pa rticular it acts like a pin-hole camera in th at it sharp ly focuses all parts of an

image. Real cameras are strictly limited in their ability to simultaneou sly focus

the images of objects placed at d ifferent distan ces in front of the lens this is a

ph ysical limitation that d epend s solely on the ratio of the d iameter of the

aperture of the cameras lens to its focal length. In photography this ratio is

called th e f-stop or f-number of a lens. This opt ical limitation is

aesthetically exploited by photograph ers wh en the foreground and / or the

background of a comp osition is d eliberately d efocused.

Real camer as are likewise un able to shar ply focus on objects in a scene if either

the camera is moving, particularly if it is hand -held, or if the scene, or par ts of

it, are in motion. To accomoda te wh at, in compu ter grap hics, is called m otionblur th e specification of a synthetic camera mu st includ e the idea of an image

being form ed ov er a finite period of time ie. as if using a shutter. The basic

virtual camera used so far has assum ed all motion is frozen by capturing an

instantaneous rep resentation of a synth etic scene.

If ph oto-realism is to be achieved in th e digital dom ain then similiar facilities

should be available to 3D comp uter gr aphics. By p roviding m echanisms to

introduce, what are in effect, digital versions of depth of field and motion

blur, RenderMan greatly im proves a designers ability to deliver ph oto-realistic

imagery. Interestingly, some real world photographic nuances are not available

in RenderMan, these includ e effects caused by,

lens flare,

optical defects such as astigmatism and chromatic aberration, and

the mu nd ane effects of dirty or scratched optics!

Because there is a slight d ifference in th e way field of vision (fov) is defined

in ph otography compared to Rend erMan, this section also provides some

examples of converting from on e system to th e other.

An Improved Camera

Overview


57/120

An Improved Camera 6 2

rman booklet

Documents