Lynn Zhou

Project 1 Overview

This project used pictures taken by Prokudin-Gorskii who attempted to create color photography by taking a sequence of black and white photos of the same scene, but with a red, green, and blue filter such that when they are combined, we obtain the original color of the scene.

My approach

The way I approached aligning the images was that after splitting the negatives into the individual blue, green, and red filtered versions, I aligned the images against the blue image by using an image pyramid. The image pyramid rescales the photos until they are 1/64 of the original size, and then manually searches over 25 possible displacements (plus or minus 5 pixels in each direction) and keeps track of the best alignment. For alignment I implemented both the SSD and NCC algorithms and found maximizing NCC more effective than minimizing SSD. To speed NCC up and make it more accurate, I also cropped 15% of the borders off the images before calculating the NCC so it focuses on aligning the middle portion and any noisy borders are not creating false positives/negatives. Without cropping these borders, many more of the images were pretty blurry.

emir.tif is still slightly blurry possibly because the scores calculated by NCC may not be representative of the actual similarity. You can kind of see that the red plate seems to be off the most. Looking at the glass-plate negatives, the red and the blue filter pictures are very different since his clothing is bright blue, so that could have caused some mis-alignment.

Results on example images

cathedral.jpg

green displacement: [2, 5]

red displacement: [3, 12]

time elapsed in seconds: 0.69

nativity.jpg

green displacement: [1, 3]

red displacement: [0, 8]

time elapsed in seconds: 0.47

settlers.jpg

green displacement: [0, 7]

red displacement: [-1, 15]

time elapsed in seconds: 0.51

monastery.jpg

green displacement: [2, -1]

red displacement: [2, 3]

time elapsed in seconds: 0.58

emir.jpg

green displacement: [17, 60]

red displacement: [14, 124]

time elapsed in seconds: 59.8

harvesters.jpg

green displacement: [17, 41]

red displacement: [23, 89]

time elapsed in seconds: 63.4

icon.jpg

green displacement: [9, 55]

red displacement: [12, 117]

time elapsed in seconds: 59.4

lady.jpg

green displacement: [29, 79]

red displacement: [37, 176]

time elapsed in seconds: 70.4

self_portrait.jpg

green displacement: [14, 52]

red displacement: [11, 111]

time elapsed in seconds: 58.4

three_generations.jpg

green displacement: [6, 43]

red displacement: [32, 87]

time elapsed in seconds: 57.7

train.jpg

green displacement: [21, 56]

red displacement: [28, 116]

time elapsed in seconds: 61.2

turkmen.jpg

green displacement: [12, 65]

red displacement: [22, 137]

time elapsed in seconds: 60.3

village.jpg

green displacement: [24, 49]

red displacement: [58, 103]

time elapsed in seconds: 60

Results on some of the Prokudin-Gorskii collection

green displacement: [5, 25]

red displacement: [7, 111]

time elapsed in seconds: 64.8

green displacement: [-6, 25]

red displacement: [-28, 74]

time elapsed in seconds: 65.1

green displacement: [-4, 55]

red displacement: [-7, 129]

time elapsed in seconds: 65.6

green displacement: [3, 16]

red displacement: [4, 42]

time elapsed in seconds: 63.8

green displacement: [2, 40]

red displacement: [-3, 82]

time elapsed in seconds: 61.6