An Improved Motion Estimation for Spatially Scalable Video Coding

In the incoming amendment of H.264/AVC for scalable video coding (SVC), inter-layer prediction is incorporated into spatial scalability. According to inter-layer motion prediction, the base-layer and enhancement-layer can use one set of motion vectors, due to the strong relationship between different layers. In this paper, an improved motion estimation scheme is proposed for gaining optimal motion vectors, aiming to balance the coding efficiency between two layers, meanwhile maintain the final coding performance. The proposed scheme performs motion estimation in both two layers simultaneously, and uses quantization step as weighting parameters, which can ensure that the coding performance can not be affected by variable QP(quantization parameters). Thus, the proposed cost function is optimized. When the cost is minimal, an optimal motion vector is obtained, and the graceful tradeoff between the two layers is also achieved. Our experimental results show that the proposed scheme can balance the coding efficiency between the two layers perfectly, and at the same time improve the total coding performance.