cathedral.jpg Green: (5, 2)
Red: (12, 3)
monastery.jpg Green: (-3, 2)
Red: (3, 2)
tobolsk.jpg Green: (3, 3)
Red: (6, 3)
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. Thus, I extracted the three color channel images, placed them on top of each other and aligned them so that they form a single RGB color image.
First, I processed and prepared the .jpg image by splitting it into three equal sections, cropping the outer 15% of the image to remove the dark or black borders and yield a more accurate alignment. I then aligned the images by exhaustively searching over the interval [-15, 15] and finding the best alignment based on Normal Cross-Correlation (NCC). A higher NCC meant a better alignment. Finally, I returned the image with the best alignment, or highest NCC.
cathedral.jpg Green: (5, 2) Red: (12, 3) |
monastery.jpg Green: (-3, 2) Red: (3, 2) |
tobolsk.jpg Green: (3, 3) Red: (6, 3) |
In order to process larger .tif images, I used image pyramids. I processed every image using the same method as I did in the single-scale alignment. Then, I rescaled the image by 0.5 , halving its size for 2 iterations. This means that the pyramid has 2 levels of 0.5 and 0.25 of the original image. I aligned the images starting from the coarsest level (level 2 , 0.25 of the original image), adding each level's displacement multiplied by 2^level to d = (dx, dy) , initialized to (0, 0) . Finally, I used d as the displacement to align the original .tif image and returned it.
church.tif Green: (48, 8) Red: (86, -8) |
emir.tif Green: (78, 44) Red: (58, 40) |
harvesters.tif Green: (90, 34) Red: (90, 14) |
icon.tif Green: (70, 34) Red: (90, 46) |
lady.tif Green: (82, 16) Red: (90, -6) |
melons.tif Green: (90, 4) Red: (90, 4) |
onion_church.tif Green: (82, 56) Red: (90, 66) |
sculpture.tif Green: (94, -30) Red: (210, -78) |
self_portrait.tif Green: (170, 80) Red: (210, 144) |
three_generations.tif Green: (130, 42) Red: (202, 30) |
train.tif Green: (114, 18) Red: (178, 94) |
I chose 4 images from the given link in the project description. I downloaded 2 as .tif files and the other 2 as .jpg files. I aligned the .tif using multiscale alignment and the rest using single-scale alignment.
custom1.tif |
Green: (70, 18)
Red: (90, 8) |
custom2.tif |
Green: (74, -24)
Red: (90, -54) |
custom3.jpg |
Green: (2, 1)
Red: (7, 1) |
custom4.jpg |
Green: (2, 2)
Red: (8, 4) |