CS294-6 Fall 2006Problem Set #1R. Bajcsy, S. Sastry, and A. YangIssued: September 6 Due: September 20 in classPrimary Reading: Chapter 1, 3, and 4 of MaSKS.Advance Reading: Appendix A of MaSKS.Note: A homework set in this class typically consists of two parts. The first part consists ofexercises from the textbook. In each set, there will be a problem with a (*) notation at the start toindicate that it is considered harder than the requirements for the course. Accordingly, the pointsfor solving this problem are reduced.The second part consists of MATLAB programming exercises. Sufficient programming skills onMATLAB and its Image Processing Toolbox are required, but you can always teach yourself u singthe documentation and demos in MATLAB. Please print out your codes, source images, andresults.Problems:1. Exercise 3.6 (10 points)2. Exercise 3.9 (10 points)3. (*) Exercise 4.7 (5 points)4. Programming Exercise 4.8 (10 points)Hint: Useful MATLAB commands: imread, imshow, im2double, mat2gray, and (of course,most useful) help your command here.5. Lab Tour and Programmi ng Exercise (10 points)• Objectives:Analyze fixed pattern noise of CCD sensors in a stereo cluster. Compare the resultsbetween two cameras in the cluster. Analyze average noise of individual pixels, averageintensity p e r column, power spectrum and the histograms (refer to [1]).• Methods:CCD camera noise has three major components: photon noise, read noise and fixedpattern noise [1]. Examine the noise by capturing 30 dark images on two differentcameras. The dark images are the response of the cameras with no access of light to theCCD (the lens is covered with a lens cover).Next, analyze photo response non-uniformity from captured 30 flat-field images. Flat-field images are captured at homogeneous illumination with the lens taken off. A diffuseris used in front of the CCD. The cam era shutter/gain is set at values near saturation.• Organization:Students should come to Professor Bajcsy’s lab at 475 Hearst Memorial Mining buildingand collect image data from a multi-camera vision system, and anaylize the cameranoises as described in the class.1The lab session will be monitored by Dr. Gregorij Kurillo [[email protected]].Three time slots will be offered: Sep 13 (Wed), 14 (Thur), and 18 (Mon). The sessionstarts at 10:00am sharp and concludes at 10:30am. If you cant make these slots pleasecontact Dr. Kurillo. A bonus lab tour will be also offered by Dr. Kurillo.References[1] G. Kamberova, R. Bajcsy, Sensor errors and the uncertainties in s tereo reconstruction, in: K.W.Bowyer, P.J. Phillips, eds., Empirical Evaluation Techniques in Computer Vision, pp. 96-116.IEEE Computer Society Press,
View Full Document