MPEG-1 super-resolution decoding for the analysis of video still images

A digital image sequence coded at low bitrate using a motion-compensated video compression standard should contain little data redundancy. However, the success of a particular super-resolution enhancement algorithm is predicted on super-resolution overlap (i.e., redundancy) of moving objects from frame-to-frame. If an MPEG-1 bitstream is coded at a relatively high bitrate (e.g., a compression ratio of 15:1), enough data redundancy exists within the bitstream to successfully perform super-resolution enhancement within the decoder. Empirical results are presented in which decoded pictures from MPEG-1 bitstreams containing both global scene transformations and independent object notion are integrated to generate Bayesian high-resolution video still (HRVS) images. It is shown that additional spatial details can be extracted by integrating several motion-compensated coded pictures, provided that a large number of subpixel-resolution overlaps-such as those captured by a reconnaissance airplane or surveillance satellite-are present among the original digitized video frames.