Multiple and Simultaneous Image Compression for Video Sequences

We propose to segment the images by performing two-level segmentation of the images: (i) the background corresponds to the stationary regions, and (ii) the foreground contains the moving objects. We use oriented 3D Gabor filters to sense that activity. With these filters we detect the moving blocks in the direction corresponding to the motion in the whole sequence. After, based on this information, we only select the moving blocks in each image. To increase the compression rate, we apply to these moving blocks the 3D vector quantization techniques described in [2].