MOTION AND ILLUMINATION VARIATION ESTIMATION USING A HIERARCHY OF MODELS - APPLICATION TO IMAGE SEQUENCE CODING

In this paper, a new image sequence coding technique which exploits the redundant information between two successive images of a sequence is presented. We propose to model the temporal variations of the images without the usual brightness constancy assumption, using motion and variations of illumination parameters. This modeling approach leads to a hierarchy of temporal variation models. To achieve an efficient region-based adaptation, a criterion that combines the quality of reconstruction and the parameter transmission cost is introduced. This hierarchy of temporal variation models is implemented using an efficient differential estimation algorithm. High compression ratios with high qualities of reconstruction far motion- and illumination-compensated sequences are obtained. This algorithm is particularly efficient when complex natural motions and illumination variations occur in the image sequences. The proposed coding technique has been validated on real image sequences. Quantitative results in terms of peak signal-to-noise ratio and compression ratios are provided. (C) 1995 academic Press, Inc.