Project 1: Images of the Russian Empire

Colorizing the Prokudin-Gorskii photo collection

In this project, I take RGB negatives and align them to create color photographs. These filter photographs were taken by Prokudin-Gorskii in the early 1900s, before the invention of color photography. Prokudin-Gorskii's vision was to take three exposures of a scene with a red, green and blue filter, and in the future, these photos would be able to be combined into one color photograph. To align the images, I first split the original jpg file containing the three negatives into thirds. Then, I align them by downscaling them to a smaller size (~100x100) and using the structural similarity to find a roughly shift. I use this estimated shift on a slightly less downscaled version of the images to find a new approximation and continue this until back to the original size. Here are the results below.

Cathedral: g-shift = (5, 2), r-shift = (12, 3)
Object 2
Church: g-shift = (25, 3), r-shift = (58, -4)
Object 3
Emir: g-shift = (50, 21), r-shift = (105, 40)
Object 4
Garden: g-shift = (59, 27), r-shift = (125, 34)
Object 5
Harvesters: g-shift = (59, 14), r-shift = (123, 12)
Object 6
Lady: g-shift = (57, 8), r-shift = (119, 12)
Object 7
Icon: g-shift = (39, 16), r-shift = (89, 23)
Object 8
Melons: g-shift = (87, 10), r-shift = (178, 12)
Object 9
Monastery: g-shift = (-3, 2), r-shift = (3, 2)
Object 10
Onion church: g-shift = (52, 24), r-shift = (108, 35)
Object 11
River: g-shift = (43, -2), r-shift = (97, 9)
Object 12
Self portrait: g-shift = (78, 28), r-shift = (175, 36)
Object 13
Three generations: g-shift = (55, 13), r-shift = (112, 9)
Object 14
Tobolsk: g-shift = (3, 2), r-shift = (6, 3)
Object 15
Train: g-shift = (39, 16), r-shift = (89, 23)

Bells and Whistles

In this section, I showcase some auto-contrasting tools. The first is rescaling intensities, which rescales the pixel values so they range from 0 to 255 rather than just min to max. The next technique is Histogram Equalization which rescales based on the cumulative distribution of pixel values so that pixels with similar values are contrasted more. The final one highlighted is Adaptive Histogram Equalization, which works by computing local histograms by different tile regions in the image and using those to enhance the contrast.

Emir (without contrast)
Emir (with rescaled intensities)
Emir (with Histogram Equalization)
Emir (with Adaptive Histogram Equalization)
River (without contrast)
River (with rescaled intensities)
River (with Histogram Equalization)
River (with Adaptive Histogram Equalizataion)