CE 261: Lab #4
Removing Artifacts by Image Segmentation

Spring 1997

Instructor: Dr. Craig M. Wittenbrink

Office: Applied Science Bldg. #309

HP Phone: (415) 857 2329
UCSC Phone: (408) 459 4099

Due: Thursday June 5. The labs will be graded prior to the final.

All labs will be demonstrated prior to the final, and writeups will be collected at that time. Lab Environment: Computer Engineering/Computer and Information Sciences educational lab. Silicon Graphics Indy's and Indigos.

Location: Applied Sciences 213

Lab Hours: Any time, but you need a keycode

Segmentation.

The lab assignment is to remove artifacts present when view morphing the Hoover Tower view images. View Morphing was the assigned work for Laboratory number 3 [1]. The required elements of the lab are to segment out the bars, walls and floor from the distance imagery. The exact method is up to your choosing, essentially programming in C or C++. No high level image processing package may be substituded for your lab assignment( matlab, mathematica Khoros, etc.)

Segmentation may also be performed by taking the frequency processing of the image, morphological processing of the image, and median filtering etc. Because of the specific colors and shapes of the bars, there are many solutions that can be used.

Artifacts adaption

To remove the artifacts of the bars, and occlusion effects, there are many approaches. One means for extracting the bars may be to threshold, and then perform a connected components analysis. The connected components analysis algorithm is described in Haralick and Shapiro, pages 28-48 [4]. Pseudo code is given for various algorithm variants. The classical algorithm, similar to the algorithm described in Gonzalez and Woods, chapter 2, can be used, Section 2.3.4, or the ``Space-Efficient Two-Pass Algorithm That Uses a Local Equivalence Table'', Section 2.3.5

Artifact Free View Morphing

The ability to segment the images provides for the chance to more properly morph and warp the images. For example, if the images can be segmented to remove the foreground information, a proper panorama can be created by simply morphing the images together, providing an unobstructed view of what would be seen from the wider perspective. Also, the bars images could be warped separately from the distance images to avoid ghosting of the bars. For full credit on the lab, addressing the complete application of view morphing, avoiding artifacts, and making subjective decisions about the appearance are to be done.

A reference implementation of view morphing may be available to assist those whose interfaces were not adequate in Lab #3.

Lab 4, Lab Write-up

You must show in your write-up the pseudo code of the algorithms used for the segmentation. The use of the segmented images to more properly view morph the imagery is also to be described in pseudo code. Please document, and comment all of your code routines, modules and files. Also describe the implementation of your program. A listing of the program showing that the program compiles, usage etc. is to be included. I recommend that you start on the lab immediately, as programming projects often take longer than expected.

References

[1] Lab 3 assignment, Craig M. Wittenbrink.

[2] ImageVision Library Programming Guide, by Jackie Neider and Eleanor Bassler, Document Number 007-1387-030, Silicon Graphics, 1993.

[3] K. Castleman. Digital Image Processing. Prentice-Hall, 1979. Second Edition 1996, Chapter 8, Geometric operations.

[4] R.M. Haralick, and L.G. Shapiro, Computer and Robot Vision, Volume I, Addison Wesley, 1992.

Copyright, Craig Wittenbrink, 1997.

craig@hpl.hp.com

Last modified Tuesday, 27-May-1997 15:47:29 PDT.