Hybrid cuckoo search based evolutionary vector quantization for image compression

Abstract

Vector quantization (VQ) is the technique of image compression that aims to find the closest codebook by training test images. Linde Buzo and Gray (LBG) algorithm is the simplest technique of VQ but doesn’t guarantee optimum codebook. So, researchers are adapting the applications of optimization techniques for optimizing the codebook. Firefly and Cuckoo search (CS) generate a near global codebook, but undergoes problem when non-availability of brighter fireflies in search space and fixed tuning parameters for cuckoo search. Hence a Hybrid Cuckoo Search (HCS) algorithm is proposed that optimizes the LBG codebook with less convergence time by taking McCulloch’s algorithm based levy flight distribution function and variant of searching parameters (mutation probability and step of the walk). McCulloch’s algorithm helps the codebook in the direction of the global codebook. The variation in the parameters of HCS prevents the algorithm from being trapped in the local optimum. Performance of HCS was tested on four benchmark functions and compared with other metaheuristic algorithms. Practically, it is observed that the Hybrid Cuckoo Search algorithm has high peak signal to noise ratio and a fitness function compared to LBG, PSO-LBG, FA-LBG and CS-LBG. The convergence time of HCS-LBG is 1.115 times better to CS-LBG.