An efficient component-separable inter layer interpolation scheme applied in scalable video coding

摘要

Recently, active research has conducted on a new emerging video coding standard, scalable video coding (SVC), which adopts a layered coding scheme to generate a multi-layered bit stream for heterogeneous environments. One of the most important features of SVC is the utilization of inter layer prediction coding, and the coding efficiency will be greatly dependent on the design of the inter layer interpolation scheme. In the current working draft, a uniform up-sampling filter is employed which aims to achieve high interpolated precision for both luma and chroma components of coded images, but this results in extremely unnecessary computational complexity which prevents it from being put to practical use. To overcome this, an efficient component-separable inter layer interpolation scheme is proposed in this paper. Based on the basic characteristic of human vision system, this proposed scheme intends to assign suitable filters for different components according to their sensitivities for human eyes. In particular, the 6-tap filter is retained for the luma component but for chroma components, much more simplified filters such as 2-tap and 4-tap filter should be used instead. Experimental results show that when compared with the original scheme, the proposed scheme significantly reduces the computational complexity by saving on coding time while holding the coded bit-rate and PSNR performance. Therefore, this scheme can be applied in the spatial interpolation of scalable video coding.