文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Fast Intra- and Inter-Prediction Mode Decision in H.264 Advanced Video Coding
侵权投诉

Fast Intra- and Inter-Prediction Mode Decision in H.264 Advanced Video Coding

Abstract- H.264/AVC, the latest video coding standard, achieves better video compression rates since it supports new features such as a large number of intra- and inter-prediction candidate modes. H.264/AVC adopts rate-distortion optimization (RDO) techniq

Among many new features, the intra- and inter-prediction techniques are recognized to be the main factors that contribute to the success of H.264/AVC. H.264/AVC employs the Lagrangian RDO method to find out the best coding mode of intra-and inter-prediction with highest coding efficiency. Figure 2 [1] shows the RDO process. RDO technique requires a lot of computations since it tests the encoding process with all possible coding modes of intra-coding, and calculates their RD costs to choose the mode having the minimum cost. The reference software of H.264/AVC, JM [15], adopts full search for both motion estimation and intra prediction. The run time percentage of each function is shown in Figure 3[20].

For inter prediction, H.264 supports tree-structured multi-block sizes in motion estimation, permitting seven modes with different block sizes (16x16, 16x8, 8x16, 8x8, 8x4, 4x8, 4x4). As shown in Figure 3, motion estimation is the most computationally intensive part. Also, The intra-prediction mode decision is very complex and the number of computing RD cost values for luma and chroma of a macroblock is 592 [5].

Therefore, the computational burden of these types of brute force-searching algorithm is far more demanding than any existing video coding algorithm.

Fast Intra- and Inter-Prediction Mode Decision in H.264 Advanced Video Coding

Fast Intra- and Inter-Prediction Mode Decision in H.264 Advanced Video Coding

Figure 2: Computation of RD cost [1].

Figure 3: Run time percentages of functional blocks in H.264/AVC baseline encoder [20].

Fast mode selection for intra- and inter-prediction is considered in this paper; which is a challenging subject in H.264/AVC.

To reduce the computational complexity, many algorithms (such as fast motion estimation, fast inter-mode prediction, and fast intra-prediction) have been proposed. For fast inter-prediction mode decision, the early termination technique [21] reduces the number of potential prediction modes. In [22], a classification method is proposed to reduce the average number of block types while maintaining the coding performance. This algorithm exploits spatial characteristics of macroblock to help the block segmentation procedure. The FMBME algorithm in [23] is composed of two main steps- early termination based on an adaptive threshold and block segmentation.

For fast intra prediction, since it is a new topic in H.264/AVC coding with respect to other standards such as MPEG-1/2/4 and H.261/H.263 and so far no previous work exists for that. Fast intra-mode decision algorithms using edge detection histogram and local edge detection are proposed in [1, 6, 7]. However, their preprocessing stages still consume a coding time to detect the edge direction and to classify it into a limited direction. The performance of those methods is about 20~30% (or 55~65%) faster than the RDO method at the cost of 2% (or 5%) extra bits. There exist fast algorithms to select the optimal intra-prediction mode using simple directional masks in [8] with saving time of 70%, and statistical-based methods in [9] with saving time of 45%. Another fast intra-mode decision scheme is proposed in [10], where the encoding speed is approximately 30% faster than that of the RDO method. A new fast intra-prediction algorithm based on macroblock properties (FIPAMP) is presented in [11]. This algorithm can achieve 10% to 40% of computation reduction while maintaining similar PSNR and bit rate performance of H.264/AVC codes. In [12], an efficient intra-prediction (EIP) algorithm based on early termination, selective computation of highly probable modes, and partial computation of the cost function is presented. Also, an improved cost function to improve the coding performance is proposed in [13].

In [14], a fast algorithm based on the local edge information obtained by calculating edge feature parameters, and sub-sampling of matching operations is presented. That method can reduce the encoding time about 26% with less than 1.4% used extra bits and no more than 0.7 dB PSNR is sacrificed.

Here, for intra prediction mode decision, we improve Pan’s method [1, 6, 7], that is eliminating the DC mode from the candidates if the direction of block is obvious, otherwise, DC mode is chosen. As an alternative method, we proposed a fast intra method using statistical properties of adjacent MBs and reference pixels with combination of the presented algorithms in Pan’s method.

上一页第2页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email