PSO image thresholding on images compressed via fuzzy transforms

We present a new multi-level image thresholding method in which a Chaotic Darwinian Particle Swarm Optimization algorithm is applied on images compressed by using Fuzzy Transforms. The method requires a partition of the pixels of the image under several thresholds which are obtained by maximizing a fuzzy entropy. The usage of compressed images produces benefits in terms of execution CPU times. In a pre-processing phase the best compression rate is found by comparing the grey level histograms of the source and compressed images. Comparisons with the classical Darwinian Particle Swarm Optimization multi-level image thresholding algorithm and other meta-heuristic algorithms are presented in terms of quality of the segmented image via PSNR and SSIM.