Hyperspectral regression lossless compression algorithm of aerospace images

Abstract

In this work, we propose an algorithm for compressing lossless hyperspectral aerospace images, which is characterized by the use of a channel-difference linear regression transformation, which significantly reduces the range of data changes and increases the degree of compression. The main idea of the proposed conversion is to form a set of pairs of correlated channels with the subsequent creation of the transformed blocks without losses using regression analysis. This analysis allows you to reduce the size of the channels of the aerospace image and convert them before compression. The transformation of the regressed channel is performed on the values of the constructed regression equation model. An important step is coding with the adapted Huffman algorithm. The obtained comparison results of the converted hyperspectral AI suggest the effectiveness of the stages of regression conversion and multi-threaded processing, showing good results in the calculation of compression algorithms.