FUZZY COLOR IMAGE ENHANCEMENT USING EVOLUTIONARY ALGORITHM

Abstract

In this study, we represent a new optimal simplified approach for image enhancement of color images using fuzzy logic and particle swarm optimization (PSO). The image exposure defined in [13] is simplified and is used to categorize the image into underexposed and overexposed regions. Objective measures like visual factors, contrast factors and fuzzy contrast defined in [13] are further removed to make the color image enhancement algorithm less complex. The hue, saturation and value (HSV) color space is utilized for the enhancement process. The hue component is kept same to preserve the original color of the image. The luminance component associated with each pixel intensity is fuzzified using gaussian membership function for underexposed as well as overexposed regions. These membership values are then modified using sigmoidal membership function to obtain the enhanced membership values and then defuzzified in order to obtain the enhanced image. The power-law transformation is used for the enhancement of the saturation component. A new objective function comprising entropy, edge information and the image exposure is introduced and optimized using PSO to learn the parameters used for the enhancement of a given image. Entropy, histogram flatness, histogram spread and tenengrad value are used for the quantitative analysis of the enhanced image. The proposed approach is evaluated using different test images that include underexposed, overexposed, mixed-exposed and low contrast images. The proposed approach is compared with other enhancement techniques available in the literature. On comparison, it is found that the proposed algorithm out performs most of the existing algorithms available in literature.