Performance comparison between two temporal wavelet lifting schemes for video coding

This paper compared the performances of the two wavelet lifting schemes when they were used to enhance temporal scalability for video coding. In the first scheme the motion vectors are estimated along double direction (forward and backward) between the two adjacent frames. However, in the second method, the motion vectors are got along one direction between the two adjacent frames. The three factors, concentricity of energy, PSNR of the reconstructed video and computation complexity were compared. Experimental results show: the concentricity of energy for the two schemes is comparable; at the same loss the PSNR of the reconstructed video for the first lifting scheme is lower about ldB than that for the latter scheme; the computation complexity of the former is higher than that of the latter on account of additional motion estimation.