Non-uniform image compression using a biologically motivated selective attention model

We propose a new non-uniform image compression algorithm using a biologically motivated selective attention model for the effective storage and transmission of natural images. The proposed selective attention model, which uses a bottom-up saliency map (SM) together with top-down reinforcement and inhibition, can generate a scan path that contains plausible interesting objects in a natural scene. The proposed non-uniform image compression method uses the SM results of the proposed selective attention model, which compresses the selected areas that are interesting and the uninteresting areas in a different way by a lossless coding algorithm and lossy compression, respectively. Experimental results show that the proposed non-uniform compression method provides a better peak signal-to-noise ratio (PSNR), but slightly decreases the compression ratio.