Improving H.264 performances by quantization of motion vectors

Abstract

The coding resources used for motion vectors (MVs) can attain quite high ratios even in the case of efficient video coders like H.264, and this can easily lead to suboptimal rate-distortion performance. In a previous paper, we proposed a new coding mode for H.264 based on the quantization of motion vectors (QMV). We only considered the case of 16x16 partitions for motion estimation and compensation. That method allowed us to obtain an improved trade-off in the resource allocation between vectors and coefficients, and to achieve better rate-distortion performances with respect to H.264. In this paper, we build on the proposed QMV coding mode, extending it to the case of macroblock partition into smaller blocks. This issue requires solving some problems mainly related to the motion vector coding. We show how this task can be performed efficiently in our framework, obtaining further improvements over the standard coding technique. INTRODUCTIONThe trade-off between allocation of coding resources to motion vectors (MV) or to transform coefficients has a major importance when it comes to efficient video coding techniques. Nevertheless, in video coding standards [1, 2], there is not much flexibility at this end: generally it is only possible to indirectly choose how the bit-rate is shared between motion and residual by selecting one among the several available coding modes for each macroblock (MB). Therefore, in several prior works, it has been noted that when a sequence is encoded at low and very low bit-rates, a large quota of resources is allocated to MVs [3]. This suggests that, in the framework of a H.264/AVC-like coder, there could be room for improvement if some new coding mode with less costly motion information is introduced, for example by quantizing the motion vectors.The quantization of MVs has been an active research topic since the mid'90 [4,5], but in these works the proposed approach was mainly a vector quantization (VQ) of MVs, with quantized vectors used both at the encoder and decoder sides (closed loop). For example in [6] a RD-optimized codebook for VQ of MVs is designed. In [7] a model based optimization of the MV precision is proposed, but here we are focused on a data-driven solution rather than a model-based one. In particular we are interested in the use of a new coding mode based on motion vector quantization.However it should be noticed that any new mode introduced into a H.264-like coder causes an increase of the coding rate, because it is more costly to signal the chosen coding mode among

Adaptive motion vector resolution with implicit signaling

“…The authors later extended their works and developed a practical scheme for H.26L encoders [3]. In a recent work [4], a H.264-based system has been presented. In all the above methods, the selection of MV resolution is based on rate-distortion (R-D) criterion.…”