Low Complexity, High Efficiency Probability Model for Hyper-spectral Image Coding

Abstract

This paper describes a low-complexity, high-efficiency lossy-to-lossless coding scheme for hyper-spectral images. Together with only a 2D wavelet transform on individual image components, the proposed scheme achieves coding performance similar to that achieved by a 3D transform strategy that adds one level of wavelet decomposition along the depth axis of the volume. The proposed schemes operates by means of a probability model for symbols emitted by the bit plane coding engine. This probability model captures the statistical behavior of hyper-spectral images with high precision. The proposed method is implemented in the core coding system of JPEG2000 reducing computational costs by 25%.