In this paper, a new fast encoding algorithm based on an efficient motion estimation (ME) process is proposed to accelerate the encoding speed of the scalable video coding standard. Through analysis of the ME process performed in the enhancement layer, we discovered that there are redundant MEs and some MEs can simply be unified at the fully overlapped search range (FOSR). In order to make the unified ME more efficient, we theoretically derive a skip criterion to determine whether the computation of rate-distortion cost can be omitted. In the proposed algorithm, the unnecessary MEs are removed and a unified ME with the skip criterion is applied in the FOSR. Simulation results show that the proposed algorithm achieves computational savings of approximately 46% without coding performance degradation when compared with the original SVC encoder.

In a scalable video coding (SVC), spatial down/up-sampling of video sequences is an essential process for spatial scalability. We propose an arbitrary-ratio spatial down-sampling method based on the type-II DCT, with which down-sampling of H.264 SVC can be efficient. Simple windowing operation in the DCT domain is efficiently embedded in the proposed down-sampling method for aliasing reduction. A fast algorithm for arbitrary-ratio down-sampling is introduced with matrix decomposition and fast DCT. It achieves 80 % reduction of computation complexity in comparison with the direct matrix calculation.

When a video sequence is compressed with rate control for a constant bit rate (CBR) channel, the visual quality usually fluctuates according to image complexity. We introduced distortion models under the CBR condition to predict distortion during rate control. The paper also proposes a control method for improvement of visual quality in H.264 CBR condition. In the experiments, we analyzed the peak signal-to-noise ratio (PSNR), the variance of the PSNR, and the average difference of the PSNR between adjacent frames for various video sequences. Our method demonstrates a more stable visual quality without severe PSNR degradation.