A Novel Region -Based Color Video Compression Technique

Abstract

This paper presents a novel video compression technique tailored to achieve high compression ratio for synthetic and thermal video sequences while maintaining a low level of complexity. The technique relies on the segmentation of each frame into a number of regions. The contours of these regions are subsequently compressed using the entropy encoding. The high compression ratio achieved by this technique can be traded for higher temporal sampling rates to allow the representation of high mobility objects. Three types of video sequences are used to evaluate the proposed technique. These are synthetic, thermal, and real video sequences. Compression Ratio (CR) as well as Peak Signal to Noise Ratio (PSNR) averaged over all frames are calculated for each type and compared to standard video compression techniques. Although the actual CR achieved is dependent on the number and size of the regions within the frame, the frame size, as well as its color representation; our results show significant improvement in CR up to twenty times the CR achieved using standard techniques. Meanwhile, the reduction in PSNR is limited. Although the relative performance of the proposed technique compared to the standard MPEG-I and MPEG-4, is maintained for all types of video sequences; the observed reconstructed video quality for synthetic and thermal videos is superior to the quality of real videos.