A New Method of Color Image Compression Based on Real-Complex GHA

Abstract

This paper proposed a new method for color image compression, for the image model, the three channels signals noted R, G and B, firstly were mapped into Greaves space, and obtained the real part and complex part for processing. And then the real/complex matrix data was segmented several sub-matrix for the training. In the real and complex domains, the data was sent to the designed Real and Complex GHA neural networks for training and extracting principle components which were forming the weights vectors w in the trained networks. The experimental results had indicated its effectiveness and feasibility.