Optimal bitstream adaptation for scalable video based on two-dimensional rate and quality models

In this paper, a two-dimensional (2D) rate model is proposed considering the joint impact of spatial (i.e., the frame size) and SNR (i.e., the quantization step) resolutions on the overall rate-distortion performance. A related 2D quality model is then proposed in terms of perceptual quality. Then the two proposed models are applied to scalable video to address the problem of optimal bitstream adaptation. Experimental results show that the proposed rate and quality models fit the actual data very well, with high coefficients of determination and small relative root mean square errors. Moreover, given the bandwidth constraint and required display resolution of the end users, the optimal combination of SNR and spatial layers that provides the highest perceptual quality can be achieved using the proposed models.