A scheme of high-compression image coding

Abstract

An image compression scheme based on the wavelet transform (WT) and human visual properties is presented. The design of the scheme is aimed to develop an image coding technique, mainly, for natural images, of high performance; this means that high compression ratio and good reconstruction quality is both an objective measure and a subjective visual effect. The technique should also be easily implementable so that it can be readily put into practical use. To fulfill these requirements, the WT is chosen to be the first basis of the technique. It is well recognized that the WT by its very nature is superior in image coding compared with those transforms traditionally used for image coding. Then a type of quantizer design based on a human visual property, known as contrast-masking, is used for quantization of the subimages at various resolution levels. With fine adjustment of the parameters for different resolution levels, this quantizer leads to good image reconstruction quality. A combined coding scheme with high coding gain, consisting of a modified RLC followed by Huffman coding, is presented in accordance with the properties of the quantized subimages. Finally, some experimental results with the Lena and Girl images are given.