Phil Harding

CSE 576 Spring 2008

Project 2

Test Sequence:

Panorama: test_seq.html

Full Res Version: stitch4.mine.jpg

The test sequence shows obvious divisions between images, most likely due to differences in exposure

and color balance when the pictures were taken.

Also this is not a spherical panorama, as there were only 4 images.

Panorama: door.html

Full Res Version: door.jpg

This panorama shot on the kaiden head from my front door worked fairly well. There is a little bit of

blurriness where images blend together and an extra banister on the far right. I do not know why

there is an extra half a picture on the right side of this image.

Panorama: yard1.html

Full Res Version: yard1.jpg

This panorama shot without a tripod has issues. It appears as though the correct motion model

was not calculated by Panorama. Perhaps a scheme that is less prone to outliers than a simple

average transformation vector would work better here.

Overall this problem seems very sensitive to problems with feature matching, therefore a good feature detector

is essential. The spherical warping and the stiching elements worked well, but if the wrong average transform is given

to the aligning program, there is little it can do, but generate images like the handheld one above.

In addition, I was not able to get the panoramas software to work. I even downloaded the ivr, jpg and html files that rick

had on the class website, and they would not work. Therefore I do not know if the links above will work.

Phil