Specular Image Capture and Evaluation for Microgloss Uniformity Measurements.

Daniel S. Hann RIT/Xerox Corp.

Advisors: Dr. Jonathan Arney RIT, Dale Mashtare Xerox Corporation

5-22-2003

2

Abstract:

The purpose of this experiment was to develop a two-dimensional specular image

capture system to quantify print microgloss uniformity for correlation of subjective

scores and objective results.

Large fov (field of view), on-axis reflection and small fov, off-axis reflection

imaging systems were used in analyzing a printed test pattern consisting of Cyan,

Magenta, Yellow, Black, Red, Green, Blue and Process Black Patches. These test

patterns have been rated for microgloss uniformity. Defects affecting the microgloss

uniformity include streaks, graininess and mottle.

Polarizers and image subtraction techniques were used to minimize the diffuse

signal within the images to be analyzed. The small fov, off-axis images were analyzed

for standard deviation of the pixel values within the color patch and maximum and

average standard deviation within the print (seven patches). The large fov, on-axis images

were analyzed for graininess of the color patch and maximum and average graininess

within the print (seven patches). Color dependency was also investigated for both

systems.

Standard deviation in the small fov system did not correlate well with the

microgloss uniformity scores. A statistically significant correlation was found in the large

fov system between maximum and average graininess and microgloss uniformity scores.

The red, yellow and green color patches had the most statistically significant signals with

R2 values of 0.98, 0.97 and 0.90 respectively (logarithmic fit) and black the lowest at

0.42.

3

Introduction: “Specular gloss is the perception by an observer of the mirror-like appearance of a

surface.” 1 Objective methods such as gloss meters measure gloss over a small area at

industry standard gloss angles providing average gloss values over the measured region.

Several illumination/capture angles are available in commercially sold glossmeters

including 20/20, 45/45, 60/60 and 75/75-degree. The illumination and capture angles are

measured from the normal. Common gloss meters do not provide the important spatial

information associated with a subjective gloss evaluation. There is no measure of gloss

variation within the measured region due to variations in the form of microgloss

structure, gloss mottle, or gloss defects. Many things including printing processes, paper

surface, toner surface and defects within the surfaces can cause these variations. These

defects can lead to color desaturation and affect perceived image quality. “Additionally,

studies have indicated that perceived differential gloss and even average gloss preference

can be significantly altered by the presence of these microgloss artifacts.”2 Currently

INCITS (International Committee for Information Technology Standards) has organized

to address standardization of many image quality metrics and test methods.3 Gloss

uniformity including microgloss defects is one of many areas which INCITS is

addressing.

“Image quality is the overall measure of success of a color printing system.”4 Microgloss uniformity is but one of many image quality metrics for which prints are

rated. Currently at Xerox Corporation microgloss uniformity is subjectively evaluated by

a team of 4-6 image quality experts visually evaluating specular gloss uniformity test

prints from many printing products currently in development. Over 100 prints have been

rated spanning many different printing technologies. Prints are rated as a means for

determining print quality. Specifications for products are based on the market. High-end

printers may have much higher quality specifications than low-end printers, for example.

This project will utilize image capture and polarization techniques to capture a 2-

dimensional specular image. These images will be analyzed for average gloss using a

Gardner 75o gloss meter, L* using a Gretag spectrophotometer, mean and standard

deviation of the pixel values using Adobe PhotoShop and Graininess using IQAF

4

software. “ IQAF (Image Quality Analysis Facility) is a proprietary software package for

determining the image quality of hard copy output from printing systems, using data

captured from a variety of input devices.” 5 The results will be correlated to subjective

microgloss uniformity scores.

Microgloss uniformity variations usually appear as graininess, mottle or streaks.

Graininess and mottle are usually not directionally oriented where streaks usually are. An

example of process caused streaks is fusing streaks. Gloss mottle is how the gloss

changes within a uniform printed or non-printed area. It is generally lower frequency

than graininess. Graininess can include micro gloss defects that are small areas of

variable gloss caused by micro defects in either the substrate, toner or other defects.

Graininess is usually smaller in size than mottle but can be high or low frequency. The

graininess algorithm used in this experiment takes the human visual transfer function into

account.

This project will not attempt to correlate mottle or streaks to subjective gloss

evaluations. Two imaging systems are used in this project. The gloss uniformity pattern

consists of 8 color patches C, M, Y, K, R, G, B and PK (Cyan, Magenta, Yellow, Black,

Red, Green, Blue and Process Black) that are 20 x 33 mm. The process black patch is not

used for the subjective evaluation. The scores for microgloss uniformity generally run

from 0 to ~ 4.5 with 4.5 being very non-uniform.

Dr. Jonathan Arney-RIT, James Michel-RIT and Klause Pollmeier of the George

Eastman House used polarizers to reduce diffuse component of images to study surface

gloss and topography of photographic prints.6,7 The image capture technique used in

those experiments utilizing polarizers will be used in this project. ImageXpert, Inc. out

of New Hampshire has developed a fixture and techniques for gloss and micro-gloss

measurements.8

Small Field of View, Off-Axis System: The first imaging system uses a camera with a small field of view. The image is

captured just off-axis from the diffusely transmitted illumination. This system should

capture defects that cause microgloss variations and lead to color desaturation and gloss

5

non-uniformity. A polarizer is used to polarize the illumination upon the test print. A

second polarizer polarizes the illumination that is reflected off the test print prior to the

camera lens. An image is captured with the polarizers in parallel with respect to each

other. A second image is taken with the polarizers perpendicular with respect to each

other. The first image contains both first and second surface reflections (specular and

diffuse). The second image should be predominately second surface (diffuse) reflections

as any light reflected from the first surface will have the same polarization state as the

incident illumination and will not pass through the second perpendicular polarizer. By

subtracting the second image from the first; a first surface reflection or specular image is

obtained. Standard deviations of the pixel values/mean of these images will be

correlated with gloss uniformity scores. This system has a small field of view. The size

of these images will not be large enough to analyze for graininess.

Large Field of View, On-Axis System: The second imaging system utilizes polarizers as in the small fov, off-axis system,

but with a larger field of view. The image is captured on axis with the image capture

angle equaling the illumination angle. The illumination is diffusely reflected. A

telecentric lens, which corrects for parallax error and has a larger field of view, will be

used. Images will be analyzed for graininess and correlated to microgloss uniformity

scores. Graininess takes into account the human visual transfer function where a score of

1-graininess unit is the minimum detectable by the human visual system.

6

Methods: Subjective Method:

The microgloss uniformity test pattern (Figure M-1.) is the test image that will be

used for these experiments. The pattern consists of color patches of Cyan, Magenta,

Yellow, Black, Red, Green, Blue and Process Black that are 33 x 20 mm (w x h). The

test pattern is subjectively rated in a neutral gray room under diffuse D50 (5000o Kelvin-

fluorescent) illumination. Observers view the test prints at various angles with respect to

the illumination angle to aid in detection of microgloss variation. These angles are not

fixed angles. The microgloss uniformity of the test print is then rated based on an

anchored scale where a score of 0 indicates very uniform gloss and scores above 4

indicate very non-uniform gloss.

Figure M-1

7

Objective Methods: The two imaging systems used in this project captured segments of the gloss

uniformity test pattern patches in axis with the short dimension of the patch (height). All

images were captured in 8-bit gray scale. Average gloss values of the color patches on

the test prints were obtained using a BYK Gardner 75o gloss meter. L* values of the

color patches on the test prints were obtained using a Gretag Spectrolino

spectrophotometer.

Small Field of View, Off-Axis System: The small fov, off-axis imaging system (Figure M-2, next page) utilizes a

transmitting diffuser and two polarizers. A Hitachi Denshi, Ltd. CCD camera with

Dazzle image-capture software was used with a Dolan-Jenner fiber Optic light source

without the fiber optic cable at full intensity. A diffuser and a fixed angle polarizer are

placed in front of the Illumination source. The diffuser produces diffuse illumination for

the image. A fixed angle polarizer is placed in front of the diffuser. A variable angle

polarizer is placed in front of the camera. The field of view for this imaging system was

approximately 2.3 x 2 mm (w x h). The resolution is approximately 2600 dpi.

An image is captured with the polarizers parallel. A second image is captured

with the variable angled polarizer at 90 degrees with respect to the fixed angle polarizer.

The specular image is found by subtracting the second image from the first. Images are

then analyzed using PhotoShop to find the mean and standard deviation of the pixel

values.

8

Illumination Source 10o Normal 20o

Diffuser transmitter Fixed Polarizer Variable Angle Polarizer Camera Computer

Figure M-2. The imaging system setup and geometry for the small fov off-axis imaging system.

Large Field of View, On-Axis System:

The large fov, on axis imaging system (Figure M-3, next page) uses a Hitachi

Denshi, Ltd. CCD camera as imaging system two with Scion image-capture software. A

Dolan-Jenner fiber Optic light source with the fiber optic cable at full intensity is used for

illumination. An incandescent illumination source (desktop lamp) is added to help in

illumination uniformity. The illumination is reflected off a diffuse reflector on to the test

pattern. A diffuser and a fixed angle polarizer are placed in front of the Illumination

source. The diffuser produces diffuse illumination for the image. A fixed angle polarizer

is placed in front of the diffuser. A variable angle polarizer is placed in front of the

camera lens. A 55 mm FL (focal length) telecentric lens with a 25mm depth of field was

used. This enabled imaging a larger area of the patch. The field of view for this imaging

system was approximately 28 x 21 mm (w x h). The resolution of this system was

approximately 580 dpi.

A first image is captured with the polarizers parallel. A second image is captured

with the variable angle polarizer at 90 degrees with respect to the fixed angle polarizer.

The specular image is found by subtracting the second image from the first. Images are

9

then analyzed for graininess using IQAF software. Graininess is calculated with

Equation M-1 (below) where G is the graininess, D is the density, )( fn

V is visual

transfer function as function of the mean density level and deviation from the mean,

)(’ fn

P is the power spectrum compensating for the aperture. The maximum graininess

of the color patches per print as well as the average are also obtained for correlation to

gloss uniformity scores.

Equation M-1. ∑ ××= −

fPVeG

n

ffnn

D )(’)(8.1

Diffuse Reflector Fixed Angle Polarizer 40o Normal 40o

Fiber Optic, Incandescent Illumination Image Variable Angle Polarizer Camera Computer Figure M-3. The imaging system set up and geometry for the large fov, on-axis imaging system.

10

Results/Discussion:

Microgloss uniformity does not depend on average gloss of the color patches on

the print (Figure R-1, below). A print could have low or high gloss but have either

uniform or non-uniform microgloss. This is dependent only on the original printed test

patterns, their gloss and their subjective microgloss uniformity ratings; not the images

captured in this project.

Figure R-1. Microgloss uniformity scores vs. Gardner 75o gloss of color patches on microgloss uniformity target.

0

0.5

1

1.5

2

2.5

3

0.00 20.00 40.00 60.00 80.00 100.00

Gardner 75 Degree Gloss

Glo

ss U

nifo

rmity

Sco

re

11

Figure R-2 (below) shows a sample of the image subtraction where the parallel-

polarized image minus the perpendicular-polarized image equals the specular image.

Figure R-2 Example of image subtraction to obtain specular image.

Small Field of View, Off-Axis System: The small fov, off-axis system had a field of view of approximately 2.3 x 2.0 mm

and a resolution of ~2600 dpi. The system utilized diffusely transmitted illumination and

polarizers to obtain a 1st surface reflection specular image.

L* values contain predominantly diffuse reflection where as the average pixel

values obtained in this imaging system should be predominantly specular reflection.

Figures R-3 and R-4 are three plots used to check validity of the imaging system for this

application. Average pixel value of the color patches does not depend on L* values

measured with Gretag spectrophotometer (Figure R-3, next page).

The average pixel value should correlate with the average gloss if this system is

capturing the specular image. Figure R-4 (next page) shows a logarithmic fit with an R2

value of 0.1357. This does not indicate that a predominately specular gloss image is

being captured. This may be due to the off-axis measurement. Table R-1 (next page)

shows the R2 for the individual color patches from Figure R-3. While some colors show

a stronger signal than others, none of them exhibit a statically significant relationship.

12

Figure R-3. Average pixel value vs. L* as measured with Gretag of color patches on gloss uniformity target for the small fov, off-axis imaging system.

Figure R-4. Average pixel value of the color patches vs. Gardner 75 degree gloss of color patches on microgloss uniformity target for the small fov, off-axis imaging system.

0.005.00

10.0015.0020.0025.00

30.0035.0040.0045.0050.00

0.00 20.00 40.00 60.00 80.00 100.00

L* (Gretag)

Ave

rage

Pix

el V

alue Blue

Cyan

Green

Black

Magenta

Red

Yellow

0.005.00

10.0015.0020.0025.0030.0035.0040.0045.0050.00

0.00 20.00 40.00 60.00 80.00 100.00

Gardner 75 Degree Gloss

Aver

age

Pixe

l Val

ue o

f Col

or

Patc

h

13

Table R-1. R2 values for logarithmic fits of the data in Figure R-4.

The Standard deviations of the pixel values within the color patches divided by

the mean pixel value of that patch were plotted against the gloss uniformity scores across

all test prints (Figure R-5, next page). A logarithmic fit was applied resulting in an R2

value of 0.2564. There was no statistical significance to this correlation. Table R-2 (next

page), shows the R2 values for the individual color patches across the print set. While

some colors show more of a signal than others they are all statistically insignificant with

Black having the maximum R2 of 0.339 and Green the minimum R2 of 0.220.

Microgloss uniformity scores can be approximated by a tent-like function where

the most non-uniform patches receive more weight than the least. Figure R-6 (page 15),

shows the correlation between the maximum and average standard deviation of the color

patches per print vs. microgloss uniformity score. The correlation is much better but still

statistically insignificant. Linear fits were applied and resulted in an R2 value of 0.5724

for the average and 0.5828 for the maximum standard deviation.

Color of Patch Logarithmic R2 FitCyan 0.337

Magenta 0.018Yellow 0.676Black 0.341Red 0.282

Green 0.112Blue 0.003

Average 0.253

14

Figure R-5. Standard deviation of patch pixel values/mean vs. the microgloss uniformity

scores for the small fov, off-axis imaging system.

Table R-2. R2 values for logarithmic fits of the data in Figure R-4

Color of Patch Logarithmic R2 FitCyan 0.239

Magenta 0.231Yellow 0.308Black 0.339Red 0.238

Green 0.220Blue 0.290

Average 0.266

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0 0.5 1 1.5 2 2.5 3

Gloss Uniformity Score

Sta

ndar

d D

evia

tion

of P

atch

P

ixel

Val

ues/

Mea

n

15

Figure R-6. Average and maximum standard deviation of color patch pixel values vs. microgloss uniformity score for the small fov, off-axis imaging system.

The small fov, off-axis imaging system provided images that when

analyzed for standard deviation, produced trends in the right direction to detect gloss

uniformity however, the correlations where statically insignificant. The off-axis

geometry used in this system may not produce a strong enough specular component.

Standard deviation does not correlate well with gloss uniformity scores. The pixel size in

this system was very small. Deviations in gloss uniformity can be on the order of

millimeters and may not have been captured in such a small field of view. This system

does not have a large enough field of view to produce images that can be analyzed for

fuser streaks, granularity, and mottle.

Averagey = 8.0712x + 5.9755

R2 = 0.5724

Maximumy = 13.951x + 6.4766

R2 = 0.5828

0.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

40.00

45.00

0 1 2 3

Gloss Uniformity Score

Ave

rage

and

Max

imum

Sta

ndar

d D

evia

tions

of P

atch

Pix

el V

alue

s

Average

Maximum

Linear (Average )

Linear (Maximum)

16

Large Fov, On-Axis System: The large fov, on-axis system had a field of view of approximately 28.0 x 21.0 (w

x h) mm and a resolution of ~580 dpi. The system utilized diffusely reflected

illumination and polarizers to obtain a 1st surface reflection specular image. A telecentric

lens, which corrects for parallax error, was used. A subset of the initial test prints set was

used for this correlation. Four prints were chosen for this correlation based on graininess

driving the gloss uniformity score.

As with the small fov, off-axis imaging system, the average pixel value does not

depend on the L* (Gretag) value of the test print (Figure R-7, next page). If the system is

capturing the specular reflection then it is expected that average pixel value of the color

patches will correlate with the Gardner 75o gloss of the color patches. Figure R-8 (next

page), shows a much better correlation than the small fov, off-axis system with a linear fit

R2 value of 0.7148. This however, is still lower than desired. This may be due to

perturbations in the paper and non-uniform illumination. There is some color

dependency in this correlation. The black patches have the poorest regression fit with an

R2 value of 0.445; the blue patches 0.706 and the others ranged from 0.898 (Green) to

0.951 (Yellow) (Table R-3, page 17) with an average of 0.826.

17

Figure R-7. Average pixel value of color patches vs. L* (Gretag) of color patches for the large fov, on-axis imaging system.

Figure R-8 Average pixel value of the color patches vs. Gardner 75 degree gloss of the color patches for the large fov, on-axis imaging system.

0.00

20.00

40.00

60.00

80.00

100.00

120.00

140.00

160.00

180.00

0.00 20.00 40.00 60.00 80.00 100.00

L* (Gretag)

Aver

age

Pixe

l val

ue

Blue

Cyan

Green

Black

Magenata

Red

Yellow

0.00

20.00

40.00

60.00

80.00

100.00

120.00

140.00

160.00

180.00

0.00 20.00 40.00 60.00 80.00 100.00

Gardner 75 Degree Gloss of Color Patches

Ave

rage

Pix

el V

alue

of C

olor

P

atch

es b

y C

olor

18

Color of Patch R2 Exponential Fit

Cyan 0.925 Magenta 0.936 Yellow 0.951 Black 0.445 Red 0.924

Green 0.898 Blue 0.706

Average 0.826

Table R-3 R2 values for logarithmic fits of the data in Figure R-8 for the large fov, on-axis imaging system.

Figure R-9 (next page) shows the correlation between graininess and microgloss

uniformity scores. The graininess value of each patch was used. A logarithmic fit was

applied across all color patches, which resulted in an R2 of 0.6162. Again this correlation

followed the expected trend but was too noisy to be statistically significant. Table R-4

(next page) shows R2 values for logarithmic regression fits for graininess of color patches

vs. gloss uniformity score by color. The black patch had the poorest regression fit with

an R2 of 0.420 however the other colors ranged from 0.773 (Blue) to 0.975 (Red) with an

average of 0.882. While this data looks promising a larger number of test prints is

necessary determine statistical significance.

19

Figure R-9. Graininess of color patches vs. gloss uniformity score for the large fov, on-axis imaging system.

Table R-4. R2 values for linear and logarithmic fit of graininess vs. gloss uniformity score as a function of color (Figure R-9).

0

2

4

6

8

10

12

14

16

0 0.5 1 1.5 2 2.5 3

Gloss Uniformity Score

Gra

inin

ess

By

Col

or

C olo r o f Pa tch R 2 Logarithm ic F itC yan 0 .885

M agenta 0 .824Ye llow 0.969B lack 0 .422R ed 0.975

G reen 0 .903B lue 0 .773

Average 0 .822

20

As mentioned with the small fov, off-axis system, the microgloss uniformity

scores are based on a tent-like function. A correlation between the average and

maximum graininess was made with the microgloss uniformity scores. The relationship

was found to be highly significant with logarithmic fit R2 values of 0.9683 and 0.9921 for

average and maximum graininess respectively (Figure R-10, below). While this data

showed good correlation the data set consisted of four test prints. A significantly larger

test set is needed to validate this method.

Figure R-10 Average and maximum graininess of color patches per print vs. microgloss uniformity score for the large fov, on-axis imaging system.

This imaging system may be capable of predicting microgloss uniformity scores

based on average and maximum graininess however the print population was very small.

Possible sources of noise for this system are illumination non-uniformities, perturbations

in the prints and subjective scoring. The first two are considered to be major

contributors. Perturbations can cause shadows in the resulting image. Solutions to this

are using a vacuum platen hold down system or mounting the images on card stock.

Maximumy = 9.2237Ln(x) + 5.2503

R2 = 0.9921

Averagey = 5.5144Ln(x) + 4.0414

R2 = 0.9683

0

2

4

6

8

10

12

14

16

0 0.5 1 1.5 2 2.5 3

Gloss Uniformity Score

Aver

age

and

Max

imum

Gra

inin

ess

of

Patc

hes

Per P

rint

Average Graininess

Maximum Graininess

Log. (MaximumGraininess)Log. (AverageGraininess)

21

Conclusions:

Small Field of View, Off-Axis System: Standard deviation did not correlate well with gloss uniformity scores in the small

fov, off-axis image capture system. Standard deviation may not correlate well with

visual ratings due to the size of the pixels not correlating well with the human visual

system qualities. A larger field of view system may pick up a graininess signal with off-

axis illumination. Field of view was considered a large source for noise. Defects leading

to the microgloss uniformity scores are large and may not be of a high enough spatial

frequency to be captured in such a small field of view.

Large Field of View, On-Axis System: Maximum and average graininess data from the large fov, on-axis image capture

view system correlated well with gloss uniformity scores. However, the data set was

limited. Illumination non-uniformity and perturbations in the test prints were considered

to be the largest sources of noise. The overall noise contribution of the subjective

microgloss uniformity scores is unknown but the standard deviation of the microgloss

uniformity scores is thought to be higher at higher scores.

22

Future Work:

1. The data set needs to be expanded to a statistically significant amount of test prints.

2. Test prints need to be mounted on card stock or vacuum platen to minimize

perturbations during imaging.

3. Illumination needs to be more uniform.

4. Investigate micro gloss uniformity scores for possible correlation with streaks/bands

and mottle metrics.

5. Investigate ASTM methods, which follow glossmeter geometries, including reduction

of optical cone angles to reduce unwanted specular reflection contributions at image

capture plane.

23

Bibliography: 1 Nadal, Maria E., Thompson, E. Ambler, New Primary Standard for SpecularGloss, Optical Technology Division, Gaitersburg, MD 20899, Vol. 72, No. 911, December 2000 2 Ng, Yee, Zeise, Eric, Mashtare, Dale, Kessler, John, Wang, Jeffrey, Kuo, Chunghui, Maggard, Eric, Mehta, Prashant, Nexpress, Xerox, Paxar, Hewlett-Packard and ImageXpert, USA Standardization of Perceptual based Gloss and Gloss Uniformity for Printing Systems (INCITS W1.1) 3 http://www.incits.org/tc_home/w11htm/glossdocreg.htm

4 Dalal, Edul N., Rasmussen, Rene’ D., Fumio, Nakaya, Crean, Peter A. and Masaaki Sato, Evaluating the Overall Image Quality of Hardcopy Output, Xerox Corporation, Webster, NY, Fuji-Xerox Co., Ltd., Ebina. Japan 5 Dalal, Edul N., Rasmussen, Rene’ D., Fumio, Nakaya, Crean, Peter A. and Masaaki Sato, Evaluating the Overall Image Quality of Hardcopy Output, Xerox Corporation, Webster, NY, Fuji-Xerox Co., Ltd., Ebina. Japan 6 Arney, J.S., Michel, James, Pollmeier, Klause, Rochester Institute of Technology, George Eastman House, Instrumental Analysis of Gloss and Micro-Gloss Variations in Printed Images , IS&T’s 2002 PICS Conference. 7 Arney, J.S., Michel, James, Pollmeier, Klause, Rochester Institute of Technology, George Eastman House, Technique for Analysis of Surface Topography of Photographic Prints by Spatial Analysis of First Surface Reflectance. 8 Kipman, Yair, Mehta, Prashant, Johnson, Kate, Wolin, Dave, ImageXpert, Inc., Nashua, New Hampshire, USA, A New Method of Measuring Gloss Mottle and Micro-Gloss Using A Line-Scan CCD-Camera Based Imaging System, NIP17 International Conference on Digital Printing Technologies