Using computer modelling in image coding

One of the central problems encountered in image coding concerns the high bit rate required for the transmission and storage of image sequences. To reduce this bit rate prediction coders are commonly used. This paper gives an overview of some of the aspects of the implementation of a semantic image coder to solve this problem for a specific class of images. A semantic image coder uses a three-dimensional model of the scene to predict the next image in an image sequence. Sophisticated computer graphics techniques are required to realistically render the resultant model and computer vision techniques are required to extract the parameters of the computer model from the given scene.<>