Improved SPIHT Algorithm Based on Associative Memory Neural Network and Human Visual System

Set partitioning in hierarchical trees (SPIHT) is considered the premier state-of-the-art algorithm in image compression, but it is not designed to explicitly consider the human visual system (HVS) characteristics. Extensive HVS research has shown that there are three perceptually significant activity blocks in an image: smooth, edge, and texture blocks. The paper uses the shape parameters of generalized Gaussian distribution to distinguish these three blocks of the image. Because the shape parameters are difficult to be estimated by general methods, the paper uses new associative memory neural network to estimate the shape parameter. The simulation results show that our algorithm gets better accuracy than other methods. The paper embeds HVS into SPIHT algorithm according to give different perceptual weights to different image blocks. The experimental results show that the peak signal to noise ratio (PSNR) and subjective visual quality of image are greatly improved after decompression.