A video encoder design combining edge-adaptive lifting and scalable block-matching with motion compensation

Abstract

This study aimed to achieve video compression by using a novel lifting-based hybrid encoder that also uses motion compensation. The proposed encoder separates video frames into temporal groups, within which certain frames am selected for producing temporal and spatial predictions over the rest of the frames. The predictions utilize spatiotemporal lifting together with motion, compensation. The combination of spatial information with temporal changes is inspired from the idea of edge-adaptive lifting, which alters prediction directions in images. A further incorporation of well-known block-matching methods with different levels was observed to improve the performance. To provide the first compression results, unpredicted frames and residues between the predicted frames and their originals were quantized and entropy was encoded. Peak signal-to-noise ratio (PSNR) measurements were used to measure the performance analysis of the proposed encoding method. Experimental results revealed that higher compression ratios were obtained as the block-matching level was increased, while the PSNR values remained within reasonable ranges. PSNR values of around 29 dB were obtained with a 1000:27 compression ratio.