Hybrid gravitational search and pattern search–based image thresholding by optimising Shannon and fuzzy entropy for image compression

Image compression is very significant process in image transmission at high data rate over a communication channel and to increase the storage capacity of storage device. Ordinary image thresholding is a class of clustering technique used for image compression because of its simplicity, robustness and accuracy but it is computationally expensive when extending for multilevel image thresholding. An attempt is made in this paper to reduce the computational time of multilevel image thresholding using hybrid gravitational search algorithm and pattern search (hGSA-PS) by optimising a criterion such as Shannon entropy or Fuzzy entropy for seeking appropriate threshold values. From literature, gravitational search algorithm (GSA) is designed to explore the global search space (exploitation), and pattern search (PS) is designed to exploit a local search space (exploration), so we hybridise the GSA and PS to achieve exploitation and exploration of search space by incorporating strengths and weakness of both, and results are compared with differential evolution, particle swarm optimisation and bat algorithm and proved better in standard deviation, peak signal-to-noise ratio (PSNR), weighted PSNR and reconstructed image quality. The performance of the proposed algorithm is found better with fuzzy entropy compared to Shannon entropy. © 2017 Informa UK Limited, trading as Taylor & Francis Group.