 |
The aim of this assignment was to align the digitized Prokudin-Gorskii glass plate images to produce a single RGB colored image. Image processing techniques were used to place the 3 color channel images acquired from Prokudin-Gorskii collection on top of each other to form the final colored image. The challenge was forming images with least visual artifacts after the automation of the process. |  |
 |
||
| Image Processing Techniques | ||
| Image alignment was initially constructed through a single scale implementation with an exhaustive search over a window displacement of 30 pixels. Due to the expense of this search, I implemented a gaussian pyramid to represent the images at multiple scales. By starting from the smallest image and going down the pyramid I updated the estimate for the shifting of the images for the alignment with least visual artifacts. Further techniques include:
1.Auto Cropping: The borders of the images were cropped before the alignment by a factor of 0.1, in order to avoid long detections due to black borders. 2.Auto Contrast: I used Matlab's 'adapthisteq' method, which implements contrast-limited adaptive histogram equalization on grayscale images. This function was used in the aligned grayscale images, before concatenating them to form the colored image. 3.White Color Balance: A gray world assumption was used in the adjustment of the white color balance. |
||
 |
