Lossy Hyperspectral Image Compression Based on Intraband Prediction and Inter-band Fractal

Abstract

Fractal encoding promising proficiency in area of picture compressing but not used at compression of hyperspectral images. The paper presents a novel and applicable copy hyperspectral image lossy compressing founded in intra-prediction fractals bandwidth and hybrid between bands. The hyper spectral color picture is divided to different groups of bandings (GOB). So, the intraband estimate is used the first banding to each one GOB, overworking the spatial relation, as the form encrypting between banding through a resident exploration procedure is used to other bands at apiece (GOB), maximizing resident likeness among two together banding. The fractals constraints is contracted with coded Exponential-Golomb coding entropies. So, progress the decrypted value, the forecast mistake and the remaining fractal transform, quantize and encoded into entropy. Experimental compression results show that our scheme can achieve a actual high peak signal-to-noise ratio (PSNR) at low-slung bit degree and achieve a medium PSNR increase taking into account the overall bit complexity encoding rates compared to other lossless compression methods. Furthermore, the classification of the accuracy of our reconstructed image is 99.75%, which is better than the original uncompressed image.