The processing and coding of digital image sequences

Summary form only given. Digital image processing and coding techniques for application to sequence of images have been examined. Of particular interest are methods applicable to videophones, teleconferencing, and TV broadcasting. One method under development is based on the idea that images are composed of contours, surrounding regions that are (in some sense) uniform. Another technique being developed is based on the extraction of the uniform regions in a sequence, by 3D (split-and-merge) region growing. In this case, the measure of uniformity within a region is the mean-squared error between the image sequence and a 3D approximating polynomial. A given region is coded using the coefficients of the polynomial of best fit and the 3D boundary of the region. The region boundaries are described using a 3D pyramidal structure. Two other methods, both still in the definition stage, are based on conjoint (spatial/spatial-frequency or spatio-temporal/spatio-temporal-frequency) image representations. The final method under consideration is based on the wavelet transform.<>