Consumer-level digital cameras observe a single value at each pixel, and full color images are reconstructed using demosaicking. We improved the effectiveness of standard demosaicking techniques by applying loopy belief propagation, an iterative relaxation technique from the probabilistic AI literature. The enhanced algorithms show significant improvement in mean-square error in both RGB and CIR color spaces.

We presented a refinement of histogram equalization which used both local and global information to remap the image greylevels. With only a small increase in computation time, we can improve contract enhancement over classical histogram equalization, while avoiding over-enhancement common with local histogram equalization.