US Patent 11991335 High-speed cell-based image compression

An example embodiment may involve obtaining an input pixel map of a digital image containing an array of a×b pixel macro-cells; classifying each of the a×b pixel macro-cells as P class cells for substantially lossless compression or Q class cells for lossy compression; creating a first intermediate pixel map representing: the P class cells as is, and the Q class cells with all zero values; creating a second intermediate pixel map representing: the P class cells with all zero values, and the Q class cells as is; encoding the first intermediate pixel map into a first output stream by using substantially lossless compression; and encoding the second intermediate pixel map into a second output stream by: downsampling the P class cells and the Q class cells therein, and serializing representations the downsampled cells.