Given the digitized Prokudin-Gorskii glass plate images, we will need to extract the three color channel images, place them on top of each other, and align them so that they form a single RGB color image. We will assume that a simple x,y translation model is sufficient for proper alignment.
For the single-scale implementation on low-res images, I ran a search over a window of possible displacements [-15,15] pixels, cropped only the center portion of the image, and scored each one using some image matching metric. This way, we would align on the more important parts of the image, rather than the noisy edges or less important parts of the image.
For the His-Res images, I ran a pyramid of height 4, scaling the image down by a factor of 2 each time.
Emir | Church | Harvesters |
---|---|---|
Green: (14, 28) Red: (62, 44) | Green: (4, 24) Red: (-4, 58) | Green: (18, 59) Red: (14,124) |
Icon | Lady | Melons |
---|---|---|
Green: (18, 40) Red: (24, 90) | Green: (8, 54) Red: (12, 144) | Green: (10, 84) Red: (14,178) |
Onion Church | Self Port | Three Gen |
---|---|---|
Green: (28, 50) Red: (38, 108) | Green: (28, 78) Red: (38, 176) | Green: (14, 50) Red: (14,110) |
Train | Workshop |
---|---|
Green: (6, 42) Red: (32, 86) | Green: (0 55) Red: (-12, 104) |
V I︠A︡snoĭ poli︠a︡ni︠e︡ | V Malorossīi | Sirenʹ |
---|---|---|
Green: (-2,40) Red: (8,96) | Green: (-6, 22) Red: (-8, 236) | Green: (-4, 48) Red: (-24, 96) |