Announcements

The final due date for the homework assignments has been extended by one week: the new due date is June 4.
For the final exam you will be allowed to bring two 8.5" by 11" sheets of paper with writing on both sides.

On HW3 problem 2(b), I forgot to mention that you need to L1-normalize each kernel, i.e. divide each filter by the sum of its absolute values. The corrected version is now posted on the main page. (Thanks Josh.)

The final exam is confirmed for Thursday, June 13, 3:00-6:00pm, in 207 Center Hall.

The little p's in HW2 problem 5(a) should be big P's. The ps/pdf files onthe web page have been updated with this correction. (Thanks to Satya and Junwen for pointing this out.)

The rows of the A matrix in the eight point algorithm are as follows:
[xr*xl, yr*xl, xl, xr*yl, yr*yl, yl, xr, yr, 1].

For the algorithm EUCLID_REC in T&V (page 166), the method for resolving the sign ambiguity is incorrect. The subparts of part 4 (a and b) should be switched. That is, E should be updated before T is updated. If they are not done in this order, the error introduced by the incorrect E will not be resolved and the algorithm will loop indefinitely. Also a helpful hint is to make sure that P and p are not confused...this will lead to errors that are very difficult to decipher. (From Josh.)

The book has a mistake (T&V page 156) in that the process for undoing the normalization is missing a transpose. The correct equation should be F_bar=(Hr)T * F * Hl. This is the way it is presented in Hartley's paper. If this transposition is left out, the translation will be in the y-direction rather than in the x-direction and the lines will be off by a little. (From Josh.)
The p_i and d in the 2nd to last line of T&V Exercise 7.6 are missing a hat and a bar, respectively. (Thanks to Josh and Bianca for spotting this.)

For those of you looking to bone up on eigenvalues and singular values, I put a link to a neat article by Cleve Moler of the Mathworks on the Matlab "eigshow" demo.
According to Studentlink, the final exam will be Thursday, June 13, 3:00-6:00pm.

In HW1, problem 7, the function "imreshape.m" should be "reshape.m". (Thanks to Satya for pointing this out.)

Please put "CSE 252" in the subject line when e-mailing me with questions about the course.

I decided to drop problem #9 (on SVD line fitting) from assignment #1. It might reappear on a later assignment, however. If you already did it, or would like to do it anyway, you can have some extra credit points for it.

The grading and course policies are now available on the main course web page; please read them carefully and contact me if you have any questions.
Please join the class mailing list by sending an email to majordomo@cs.ucsd.edu with the body of the message as follows:
subscribe cse252 my_email@something.something

The third row of the camera projection matrix I wrote on the board today was incorrect: please see T&V p. 39 for the correct entries (modulo a sign factor).
Handout on Matlab posted on course web page. This is not a required assignment, but it is recommended if you have not used Matlab before.