Images of the Russian Empire
CS194-26 Fall 2021
Spencer Hamilton
Introduction
The goal of this assignment is to take the digitized Prokudin-Gorskii glass plate images and, using image processing techniques, automatically produce a color image with as few visual artifacts as possible. This is done by extracting the three color channel images, placing them on top of each other, and aligning them so that they form a single RGB color image.
Low-Resolution Implementation
For low resolution images, I was able to exhaustively search for the proper displacements in a timely fashion. A nested for-loop calculates the MSE between the red or green frame with the blue frame in order to find the best vertical and horizontal displacements. The photographs are then cropped and stacked on top of one another to form a color image.
Cathedral
r: <12,3> | g:<5, 2>
Monastery
r: <3,2> | g:<-3, 2>
Tobolsk
r: <6,3> | g:<3, 3>
High Resolution Alignment
For high resolution images, I used an image pyramid to find smaller windows of alignment based on downscaled versions of the images. With the smaller window from the lower resolution image, I used the same nested for-loop with MSE to find the proper alignment. I aligned red and green to the blue channel which worked well for all but the emir.tif image, which failed to properly align the red channel.
Church
r: <56,0> | g: <16, 0>
Emir
r: <93,-309> | g:<49, 24>
Harvesters
r: <124,13> | g:<59, 16>
Logo
r: <89,23> | g:<40, 17>
Lady
r: <113,11> | g:<47, 8>
Melons
r: <178,12> | g:<82, 9>
Onion Church
r: <108,36> | g:<51, 26>
Self Portrait
r: <176,36> | g:<78, 28>
Three Generations
r: <111,10> | g:<53, 14>
Train
r: <87,31> | g:<42, 5>
Workshop
r: <104,-12> | g:<52, 0>
Aligning to Green Channel
By keeping the same high resolution alignment method and instead aligning to the green channel, I was able to produce a better color image of emir.tif.
Emir
r: <57,17> | b:<-49, -24>
Additional Pictures from Prokudin-Gorskii Collection
These are more photos from the Prokudin-Gorskii Collection processed with the same green channel alignment as above.
Cornflower
r: <29,6> | b:<-22, -15>
Waterfall
r: <81,-7> | b:<-30, 7>
Planing Machine
r: <52,-6> | b:<-36, -2>
Images of the Russian Empire
CS194-26 Fall 2021
Spencer Hamilton
Introduction
The goal of this assignment is to take the digitized Prokudin-Gorskii glass plate images and, using image processing techniques, automatically produce a color image with as few visual artifacts as possible. This is done by extracting the three color channel images, placing them on top of each other, and aligning them so that they form a single RGB color image.
Low-Resolution Implementation
For low resolution images, I was able to exhaustively search for the proper displacements in a timely fashion. A nested for-loop calculates the MSE between the red or green frame with the blue frame in order to find the best vertical and horizontal displacements. The photographs are then cropped and stacked on top of one another to form a color image.
Cathedral
Monastery
Tobolsk
High Resolution Alignment
For high resolution images, I used an image pyramid to find smaller windows of alignment based on downscaled versions of the images. With the smaller window from the lower resolution image, I used the same nested for-loop with MSE to find the proper alignment. I aligned red and green to the blue channel which worked well for all but the emir.tif image, which failed to properly align the red channel.
Church
Emir
Harvesters
Logo
Lady
Melons
Onion Church
Self Portrait
Three Generations
Train
Workshop
Aligning to Green Channel
By keeping the same high resolution alignment method and instead aligning to the green channel, I was able to produce a better color image of emir.tif.
Emir
Additional Pictures from Prokudin-Gorskii Collection
These are more photos from the Prokudin-Gorskii Collection processed with the same green channel alignment as above.
Cornflower
Waterfall
Planing Machine