CSE 455 Fall 2012 - Project #2 Feature Matching / Panoramas Tim Jang & Jesse Garrison

Feature Detection and Matching

The following shows the original images, the Harris operator images, and the ROC curve for our feature matching implementation. (The Harris operator images were very dark, so the values were scaled by 85).

Yosemite Image Results

Yosemite1 harris image

Graffiti Image Results

graffiti-roc

Panorama Mosaic Stitching

What Worked and Didn't

We've found that the results of our panorama are satisfactory. The positions of features match up as expected without any obvious artifacts. Very busy parts of the images (see the HUB panorama's right side) do come out a little blurry when blended. The exposure levels between the images take away from the overall effect, but we weren't able to correct it in the allotted time.