Motion vector concealment in temporal scalability for scalable video coding

In scalable video coding, temporal scalability adopts the inter layer prediction coding structure to increase the coding efficiency. However, any lost single bit in base layer or lower enhancement layer may result in the loss one or more frames and serious visual quality degradation in the decoded image at the receiver. To preserve good image quality, the error concealment algorithm is necessary for scalable video transmission. Some error concealment techniques adopt the motion vector correlation of inter layer in temporal scalability to conceal the damaged motion vector. As the GOP size increases, the inter layer correlation becomes weaker, thus the concealed image quality is decreases. This study proposes two methods to solve this problem. The first method uses the adoptive weighting of motion vectors in forward and backward reference frames to increase the accuracy of motion vector recovery. The other method employs the polynomial function to describe the motion vector correlation between the inter layer and the neighboring blocks. The experimental results show that these methods are effective to conceal the lost block. The concealment efficiency of the polynomial function of motion vector correlation is better than other methods 3.05-2.16 dB when GOP size increases.