Band reordering heuristics for lossless satellite image compression with 3D-CALIC and CCSDS

摘要

Remote sensing satellite images are used widely in space imaging applications as they collect significant information of ground objects through capturing the ground surface in immense wavelength bands. The size of these images is typically enormous in quantity due to the bulky number of capturing wavelengths. The images need to transmit to the ground from the sensors for a specific application. Thus, the efficient compression techniques are required to fit the available bandwidth for reducing the transmission time. The data in the images are usually redundant spatially, spectrally and temporally which give an ample opportunity to compress the images in various domains. Most importantly, the data features have a strong correlation in the separate spectral area. As a result, the similarity-based band reordering strategy is used to increase the compression performance in comparison to the image having natural band order. However, finding the optimal band reordering is still a computationally challenging problem. In this paper, three different methods namely Band Reordering based on Consecutive Continuity Breakdown Heuristics (BRCCBH), Band Reordering based on Weighted-Correlation Heuristic (BRWCH) and Segmented BRCCBH have been proposed for the compression of multispectral, hyperspectral and hyper spectral sounder data. The presented methods are different on the number and type of heuristics used for obtaining the optimal band reordering. The performances of the proposed band reordering methods are tested using CCSDS 123 lossless predictor and lossless 3D-CALIC. The experimental results show the significant improvement on compression performance by using the proposed band ordering techniques for different types of real multispectral data (3-5% using CCSDS and 2-3% using 3D-CALIC), hyperspectral data (0.2-0.7% using CCSDS and 0.8-1% using 3D-CALIC) and hyperspectral sounder data (5.5-7% using CCSDS and 4-5% using 3D-CALIC). (C) 2019 Elsevier Inc. All rights reserved.