Image Coding Based on Contour Models

The advent of the Integrated Services Digital Network (ISDN) has resulted in a resurgence of research activity in the field of image coding. The target is to transmit real time image sequences over a number (< 30) of parallel 64kbit/s channels in order to implement fast document transfer, video telephone facilities and eventually video teleconferencing. Digital images typically occupy a large amount of memory. To transmit an uncompressed sequence of images would require enormous bit rates of the order of 2 M bit/s. The need for efficient image coding strategies cannot therefore be overstated. Existing image coding techniques exploit the mathematical redundancies in the data set. They treat the image simply as a set of numbers making no use of image analysis techniques. Subsequently their error measures are more concerned with mathematical differences than perceptual significance.