putting motion into the image retrieval interface defining the colors of 3d objects elise lewis...

19
Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Upload: jada-bentley

Post on 27-Mar-2015

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface

Defining the colors of 3D objects

Elise Lewis

University of North Texas

Page 2: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Overview

Introduction Background

Retrieval issues-CBIR

Assumptions 2D vs. 3D

Study Conclusions Future Research

Page 3: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Introduction

Images are expected Automated retrieval systems have been

implemented for images 3D objects bring unique challenges to

retrieval systems Methodology is needed to study 3D

objects

Page 4: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Background

Content-based image retrieval (CBIR) Automatically extracted Feature-based query classes Color space

Histogram RGB color space

3D objects Ability to rotate and zoom

Provides a 360° view of the object

Page 5: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Assumptions and previous research

Previous research explores CBIR systems with 2D images

Little research on 3D objects and retrieval systems Take prior research and test with attributes of

3D objects Develop a methodology to measure the

differences and similarities between 2D and 3D images-Are they the same?

Page 6: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Study

How much of a difference occurs in RGB values given different views of an object?

Front view 6 views (front, rear, top, bottom, left, right)

Software defined views N=10

Viewed on web Courtesy of Arius 3D (www.arius3d.com)

3 color channels (Red, Green, Blue)

Page 7: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Image ViewsFront*

Rear

Top

Bottom

Left

Right

Page 8: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

3D objects

Page 9: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

The Histogram

Page 10: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Largest Difference in Level Distribution-How much of a color is present?

Butterfly

0153045607590

105120135150165180195210225240255

Mean SD

Le

ve

ls

Front Rear Top Bottom Left RightFront Rear Top Bottom Left RightFront Rear Top Bottom Left RightFront Rear Top Bottom Left Right

232.17

108.49

Page 11: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Largest Difference in Level Distribution-Front/Top View

Page 12: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Smallest Difference in Level Distribution

Cunieform

0153045607590

105120135150165180195210225240255

Mean SD

Le

ve

ls

Front Rear Top Bottom Left RightFront Rear Top Bottom Left RightFront Rear Top Bottom Left RightFront Rear Top Bottom Left Right

121.1

121.2

Page 13: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Smallest Difference in Level Distribution-Front/Rear

Page 14: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Largest Difference in Spread-How much of color range is present?

Arrow

0153045607590

105120135150165180195210225240255

Mean SD

Levels

Front Rear Top Bottom Left RightFront Rear Top Bottom Left RightFront Rear Top Bottom Left RightFront Rear Top Bottom Left Right

100.002

59.54

Page 15: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Largest Difference in Spread-How much of color range is present?

Page 16: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Conclusions

Views change the levels of RGB Views change the range of color Complementary views (i.e. top-bottom) do not

have same mean or SD Greatest differences occur between objects with

large surface areas versus small surface areas Depth of detail needs to be defined

How important are the shades of a color? Information needs of a browser vs. researcher

Page 17: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Limitations and Future Research

Use different color space HSV L*a*b

More images from different domains Wide variety of color-Art Detailed color-Botany

Test algorithms for weighting and combining views and values

Page 18: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

References Curtin, D. P., (2003). Editing your images: Understanding Histograms. Retrieved

from the Shortcourses Website: http://www.shortcourses.co/editing/edit-14.htm.

Gudivada, V.N., Raghavana, V.V., (1995). Content-Based Image Retrieval Systems. IEEE, 18-23.

Konstantindis, K., Gasteratos, A., and Adndreadis, I., (2005). Image retrieval based on fuzzy color histogram processing. Optics

Communications,(248), 4-6, 375-386 Lee, S. M., Xin, J., H., and Westland, S., (2005).Evaluation of image similarities

by histogram intersection. Color Research & Applications, (30), 4, 265-274

Reichmann, M., (2005). Understanding Histograms. Retrieved from the Luminous Landscape website:

http://www.luminous-landscape.com/tutorials/understandingseries

Page 19: Putting Motion into the Image Retrieval Interface Defining the colors of 3D objects Elise Lewis University of North Texas

Putting Motion into the Image Retrieval Interface ASIS&T Annual Conference2005

Thank You!

Questions, suggestions or comments?

Elise Lewis

[email protected]