Content-based indexing of images by coding levels defined as a function of reduced entropy

摘要

A video image indexing system includes a method for identifying categories in compressed digital data for improved retrieval of video images. An encoder receives a frame and digitizes the image data while a sub-band coder codes each digitized frame at a plurality of transform levels to derive categorized data segments. Data segments are stored and then retrieved according to minimum entropy of Eigenvalues in a first transform level. Preferably, the compression is improved by creating a residual image of each frame by subtracting a predicted frame image from the actual frame image. This compression of data enhances the recreation of lossless video images at the receiver.