|
This page displays the results of Colorizing the Prokudin-Gorskii photo collection. This is project 1 in 15-463. The code written in Matlab can use either Normalized Cross Correlation (NCC) or Sum Squared Difference (SSD) in aligning images. In addition, the code can cut off borders of images to disregard noise and uses a inside image crop to compare only relavent portions of the image. The code can also use the pyramid to handle large images. The pyramid scheme procides for fast alignment but will be slow if memory is low (no real way to avoid this problem for large images). The following examples are results of aligning an combining the photos together. The code as turned in uses the pyramid scheme with NCC, cutting off 1/12 of the borders on each side, and uses inside image cropping. These settings can be changed by modifying global variables in the code. To get a good image the user may need to tweek one or two settings due to noise. |
|