IMAGE ENHANCEMENT USING EVOLUTIONARY COMPUTING

Abstract

A new method for the enhancement of color images is presented which uses the fuzzy logic and differential evolution. With the application of an objective measure known as exposure the image is divided into underexposed and overexposed regions. The V component or the luminance component of the HSV color space is exploited for the enhancement process. However, the hue component is kept intact so that the color constitution of the original image remains the same. The under exposed and the over exposed regions are enhanced using two different schemes. The enhancement of the underexposed region is done using a parametric sigmoid function whereas the enhancement of the overexposed region is done using power law transformation. For good enhancement results, the appropriate values of the parameters of the sigmoid function and the gamma used in power law transformation function are required. These are optimized using differential evolution. Moreover, two separate membership functions are used to characterize the underexposed and the over exposed regions of the image, i.e., Gaussian membership for the underexposed areas and triangular membership for the overexposed areas. This method becomes universal for implementation upon all types of contrast degradations because of the use of separate membership functions and enhancement operators for the two regions. The selfadaptation of the parameters, based on differential evolution, makes the whole method automatic.