Hyperspectral image coding using spectral prediction modelling in HEVC coding framework

A novel coding framework using Spectral Prediction Modelling (SPM) in the latest High Efficiency Video Coding (HEVC) framework for Hyperspectral (HS) images is proposed in this paper. A HS image presents a wealth of data where every pixel is considered as a vector in a multidimensional space. By quantitative comparison and analysis of pixel vector's distribution along spectral bands, we can conclude that modelling can solve the distribution and correlation of pixel vectors in certain bands. To exploit the distribution of known pixel vector, we estimate the current spectral band by extracting an instant spectral band of the HS image using Gaussian mixture-based modelling using the correlation of previous bands. The estimated current spectral band named as the common informatics wavelength (CIW) image is used as the additional reference image to encode the residual of the current band and the CIW or previous band using the HEVC coding framework. Every spectral bands of HS image is treated like they are individual frames of a video. In this paper, we compare the proposed method with mainstream HS encoders including JPEG2000, JPEG, PCA-DCT, HEVC-intra, Set Partitioning in Hierarchical Trees three dimensional (SPIHT-3D) and Adaptively Scanned Wavelet Difference Reduction (ASWDR). The experimental results fully justified by different types of HS dataset. The performance of the proposed method has outperformed the other encoding algorithms in terms of rate-distortion performance of HS image compression.