Camouflage evolution simulation with Genetic algorithm

http://simulations.narod.ru/

run camouflage_ga.m

See how it works here:
http://www.youtube.com/watch?v=MGdDRJlMRbY

There are 5000 images 4x4 size (population) over background that is changed from time to time. Each image has 16 color levels in each of R G B channel. Background is also 4x4 image repeated. Fitness is 1/(1+ds) where ds mean difference between a image and background image. For crossover is used method when child is random part of one parent and another part from second parent. It is horizontal dividing to the parts. Elitism applied when best image keep unchanged to next generation. There are 2 king of mutation. Absolute mutation when some pixels get random colors. Relative mutation when some pixels get random increments to colors. Best image is on first row and last column. Worst image is second row and last column. Also shown some another 7 images. Rest 4991 images are not shown. This is Matlab program.