Low bit-rate video coding using 2-D adaptive sampling motion compensation

This paper presents a low bit-rate video coding algorithm based on variable block size motion estimation model to increase the prediction accuracy. To optimize the usage of motion vectors, a 2-D adaptive sampling structures is employed for merging those homogeneous motion fields together with minimum increase of prediction error. The motion compensated prediction error is also encoded adaptively using the hybrid VQ/DCT approach to cope with the motion field segmentation information of the current encoding frame. Experimental results show that the proposed scheme can better utilize the limited bit budget for motion compensation than classical block matching techniques and those annoying blocking artifacts can be alleviated.