Hyperspectral image compression with modified 3D SPECK

Abstract

Hyperspectral image consist of a set of contiguous images bands collected by a hyperspectral sensor. The large amount of data of hyperspectral images emphasizes the importance of efficient compression for storage and transmission. This paper proposes the simplified version of the three dimensional Set Partitioning Embedded bloCK (3D SPECK) algorithm for lossy compression of hyperspectral image. A three dimensional discrete wavelet transform (3D DWT) can fully exploit the interband correlation in a volumetric block. This provides the obvious way to process data such as with hyperspectral images. The 3D structure of the SPECK algorithm will suitably extend the exploitation of interband dependence and correlation. The proposed algorithm maintains the 3D SPECK algorithm with modification of the implementation without using any list to store the significant information. This would reduce the memory requirement of the embedded coding algorithm and improve the coding time to compress the images.