Review of compression approaches for optical remote-sensing images from post-processing perspective

Image compression is crucial for storing remote sensing data in space-constrained on-orbit systems. Current methods primarily follow two directions. The first combines wavelet transforms with post-processing, leveraging wavelets' sparse representation for multi-resolution analysis and efficient hardware implementation. The second employs compressed sensing (CS), which shifts complexity to the decoder; its integration with deep learning has advanced the field. Deep learning itself is emerging as a third framework, using joint encoding-decoding to overcome the limitations of fixed transform bases. This paper outlines these methods' principles, configurations and developments.