A multi-predictor based lossless ARGB texture compression algorithm and FPGA implementation

Abstract

. In the fields of GPU and AI chip design, the frequent read and write operations on the color buffer data (ARGB), which are intensive in graphical and image access, significantly impact performance. There is a need for applications that require random access and only read small images once. To address this situation, this paper proposes an algorithm with lower modeling complexity, yet achieving near-complex implementation results, along with its FPGA implementation method. Through actual testing on multiple images, the average lossless compression rate reached 40.3%. With hardware acceleration, the execution efficiency of the algorithm was further improved, ensuring both compression rate and speed, thus confirming the effectiveness of the algorithm.