Hybrid Codebook Optimization Technique for Vector Quantization to Preserve the Quality of the Decompressed Image

Most commonly, an ideal codebook for vector quantization (VQ) based image compression is generated using the Linde Buzo Gray (LBG) method. However, the PSNR values are found to be lower when using the LBG technique. To obtain a near-global optimal codebook many naturally inspired metaheuristic algorithms were used recently, and the search procedure used by them follows the Gaussian probability distribution. Cuckoo search (CS), a metaheuristic optimization method, can be used to optimize the codebook developed by the LBG algorithm. The CS algorithm uses a levy flight distribution function, realized by Mantegna's algorithm, rather than a Gaussian distribution. But the cuckoo search, on the other hand, takes more convergence time. So, to reduce the convergence time, the CS is hybridized with the Bat algorithm in this work. The proposed hybrid optimization technique is called the Hybrid Cuckoo- Bat algorithm (HCUBAT). In this approach, the cuckoo and bat algorithms form the cascaded structure, and the convergence time is balanced by dividing the total available iteration equally among both algorithms for optimization. We have evaluated the efficiency of the proposed methodology for 100 iterations on the 2018 BraTS dataset image and compared its performance to that of the Cuckoo Search algorithm (CSA) alone over 100 iterations. A 5-11% and 0.5-4.55 % improvement in PSNR was achieved using the hybrid technique in comparison with LBG and LBG-CS.