Project 1 Writeup
Soumya Ghosh (sghosh)
5 Feb, 2010
A Gaussian Pyramid for each image was computed and image plates were first matched at the coarsest level and then on to the finer levels.
I used the canny edge detector to extract edges from each plate,
and then computed the sum of squared difference between these edge plates, to find the optimal alignment. The motivation behind matching
edges was that all three color plates seemed to have similar discontinuities even though the actual intensities of the plates seemed to vary.
- Extra Credit Extension #1 Automatic Cropping - The images were cropped by finding the strongest edges in the image plates. These edges generally seemed to correspond to the
image borders. Only the region bounded by these images were retained as the final image.
- Extra Credit Extention #2 Contrast Stretching - A standard linear stretching was applied to the cropped image.
- Extra Credit Extention #3 Automatic White Balance - Gray world assumption.
Results Images