Optimizing a random system of cascaded video processing modules by parallel evolution modeling

Video processing algorithms tend to improve over time in terms of image quality while increasing in implementation complexity. Generally, video algorithms are developed and evaluated in isolation from the video processing system of which they will be a part, in a consumer product. The final image quality obtained by that system, however, strongly depends on the interaction of its constituent algorithms. Current methods for optimizing the overall image quality are ad-hoc, time consuming and do not guarantee the best possible result. We propose a rapid and reliable method for fine-tuning composite video processing systems based on genetic algorithms (GA). The GA method evolves toward a system configuration that gives the best image quality, driven by an objective video quality metric