Evolutionary Contrast Stretching and Detail Enhancement of Satellite Images

In this paper we present a method of evolving the contrast characteristics of a satellite image by employing a real-coded genetic algorithm with subjective fitness. The contrast properties of an image are given by its contrast-stretching curve, and the task of the Genetic Algorithm is to evolve this curve in order to find the best shape according to some subjective fitness. The subjective fitness criterion is the contrast enhancement and detail enhancement in specific regions of an image. We introduce a new crossover operator, called Gaussian Uniform Crossover (GUX), that performs a better mixing of genetic material, an operator that is similar in spirit to the uniform crossover operator employed in binary-coded Genetic Algorithms. This new operator has proven to be effective in our contrast enhancement problem. The test image employed in our experiments is a gray scale NOAA satellite image of Cape Code New England.