CS194-26 Project 1: Colorizing the Prokudin-Gorskii photo collection
Shirley Liu
Introduction
For this project, we had to use image processing techniques to produce color images out of the Prokudin-Gorskii photo collection where he used three different filters, red, green, and blue, to capture black-and-white photos of the same scene.
Method
First, the image was cropped in thirds to get the blue, red, and green channel of the photo. Then, using the SSD metric, find the best possible offset of the portion of the image of length 40. For the larger tif files, this ssd method was applied to each level of the image pyramid, scaling down by two everytime.
Bells and Whistles
After implementing the naive algortihm, the image was still not aligned very well for some of the photos, so I added edge detection using the Sobel filter and this helped tremendously. Then, I cropped the image by a set amount to remove the borders. I could not figure out how to do automatic cropping. The pictures on the right are after bells and whistles and the displacement values are those of the bells and whistles.
cathedral.jpg
green: (5, 2)
red: (12, 3)
emir.tif
green: (48, 23)
red: (107, 41)
harvesters.tif
green: (59, 18)
red: (123, 15)
icon.tif
green: (41, 18)
red: (89, 23)
lady.tif
green: (52, 7)
red: (111, 10)
melons.tif
green: (83, 10)
red: (180, 13)
monastery.jpg
green: (-3, 2)
red: (3, 2)
onion_church.tif
green: (49, 26)
red: (108, 37)
self_portrait.jtif
green: (77, 29)
red: (175, 37)
three_generations.tif
green: (49, 15)
red: (109, 12)
tobolsk.jpg
green: (42, 6)
red: (85, 32)
train.tif
green: (42, 6)
red: (85, 32)
village.tif
green: (64, 13)
red: (137, 23)
workshop.tif
green: (53, 0)
red: (106, -12)