Interframe Adaptive Data Compression Techniques for Images.

Abstract : A method of interframe motion measurement and compensation, based on approximation of piecewise linear translation of small rectangular areas, is presented. This method significantly improves the temporal correlation and permits reduction of sampling rate along temporal axis. A technique of visual characterization of interframe characteristics, based on temporal cross-sections, is described. Some commonly used statistical models for prediction of the variances of transform coefficients, in intraframe and interframe transform coding, are compared. Results show that measured statistics in transform domain result in 2 to 4 dB improvement over commonly used separable covariance models. Interframe transform and hybrid coding schemes are compared against intraframe transform coding and some simple interframe predictive coding schemes. Also, distortion-rate curves for hybrid coding, based on models of interframe motion have been plotted. The applications of transform and hybrid coding to biomedical x-ray images has been considered. A method for joint optimization of source coding and channel coding for PCM transmission over noisy channels is presented. The performance of several transforms has been compared for some commonly used intraframe nonseparable covariance models. (Author)