Motion-compensated predictive coding of image sequences : analysis and evaluation

• Main Author: O'Shaughnessy, Richard.
• Format: Others
• Language: en
• Published: McGill University 1985
• Subjects: Image transmission; Coding theory; Image processing
• Online Access: http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=72065

In this study, motion-compensated predictive coding of monochrome image sequences was investigated. The goals were to evaluate the performance limits of this coding method, to determine its relative advantages in comparison to other predictive coding techniques not compensated for motion, and to identify the factors which limit further improvements in performance. A motion-compensated prediction algorithm which estimates displacements only from previously-transmitted data was presented, and was shown to be nearly optimal. At low distortion levels, an adaptive intra-inter-frame predictor not compensated for motion was found to be almost as effective as motion-compensated prediction. At higher distortion levels, however, motion-compensated predictive coding becomes more attractive. The main factors which limit the performance of motion-compensated predictors were identified as: input noise, spatial interpolation errors, and the local nature of the displacement estimation methods. Methods to control these limiting factors were suggested.