Fast H.264/SVC inter-frame and inter-layer mode decisions based on motion activity

Abstract

In the SVC extension of H.264, motion compensation of the enhancement layer is performed in both the same resolution layer and the corresponding lower layer, which results in greatly increased computational complexity. A fast inter-frame and inter-layer mode selection algorithm utilising the motion activity in the key frames of the video sequence is proposed. Statistics obtained empirically reveal that a macroblock with slow movement is more likely to be best matched by one in the same resolution layer. However, for a macroblock with fast movement, motion estimation between layers is required. This a priori knowledge forms the basis for the mode selection method proposed. Simulation results show that the algorithm can reduce the encoding time by up to 40%, with negligible degradation in rate distortion.Introduction: The Joint Video Team of the ITU-T Video Coding Experts Group (VCEG) and the ISO/IEC MPEG has standardised a scalable video coding (SVC) extension of the H.264/AVC standard. SVC allows a partial bit-stream to be transmitted and decoded with lower temporal or spatial resolution or reduced quality, depending on device capabilities, network quality and user demand [1].For spatial scalability, the encoder performs base layer coding on a low resolution version of the video sequence using the standard H.264/AVC encoder, and enhancement layer coding on a higher resolution version of the sequence. SVC introduces many new features to improve the coding efficiency and robustness, including inter-layer prediction between base layer and enhancement layers. However, these features result in greatly increased computational complexity. The cost of mode selection is much higher than in H.264/AVC. Some fast mode selection methods for SVC have been proposed [2, 3]. They either speed up the encoding process for single layer coding or may result in rate distortion (RD) degradation if the motion patterns between layers are not similar. In [1], the authors point out that, even if the reconstructed lower layer is essentially lossless, the data may not be optimum for inter-layer prediction. For some video sequences with slow motion and high picture detail, temporal prediction generally achieves a better approximation than the up-sampled lower layer reconstruction. In this Letter, we propose an inter-layer and inter-frame mode selection algorithm using key frame information. Simulation results show that the algorithm can reduce coding time by up to 40% with insignificant bit rate increase or PSNR degradation.

Low Complexity for Scalable Video Coding Extension of H.264 based on the Complexity of Video

“…The BMA frames are divided into nonoverlapping blocks as discussed in the previous section, and each block is compared with its counterpart in the previous frame [15][16]. To find an area that is similar.…”