Image processing using genetic algorithm

Question

i want to see application of GA in image processing. I was not getting how to write the fitness function for any image. Say I have a image in grey scale,which is not very clear, now if I keep matching my generated solution with this picture value then how is my solution going to be better like better clearity increased contrast/brightness etc.

Answer 1

My favorite example of Genetic Algorithms in image processing is:

http://rogeralsing.com/2008/12/07/genetic-programming-evolution-of-mona-lisa/

It evolves random shapes into something that resembles the Mona Lisa. You can actually try it and play with it here:

http://www.nihilogic.dk/labs/evolving-images/

Your question isn't very clear. One thing that is extremely important with a genetic algorithm though, is to measure how 'good' it is. In the Mona Lisa example, goodness can be measured relatively easily, it's the proximity to the actual Mona Lisa. In your case, it sounds like you are trying to improve a pictures quality somehow with no way of measuring it's 'goodness'.

You should build a function that evaluates the pool of solutions and scores each one. Then breed the good quality solutions with each other. Crucial to all this though, really is, writing an evaluation/scoring function so you know which ones are good.

For aerodynamics, a pool of shapes can be scored on weight, drag, lift etc. For network design, layouts can be scored on latency, physical cost, interference etc. If you explain your problem in more detail we could probably help you come up with a set of metrics to score your solutions on.