DIGITAL IMAGE COMPRESSION BASED ON VISUAL-PERCEPTION AND SCENE PROPERTIES

Digital compression techniques have made impressive progress in recent years. Frequently, a reader of a paper on compression is left with th e impression that one is getting something for nothing; however, compr ession can only be achieved by leaving out unnecessary information abo ut the image. A compression ratio of 20:1 simply means that 95% of the information in the original image has been eliminated. There are only two types of information that can be removed without seeing degradati on in image quality: information that can be accurately predicted or i nformation that the human visual system cannot see. This article will discuss present compression techniques in terms of these two factors a nd show how even further compression can be obtained by developing a c omplete understanding of scene properties and those of visual percepti on. From an analysis based on present knowledge of visual perception a nd scene statistics, compression ratios in excess of 50:1 should be ac hievable without perceptible degradation.