基于HEVC的低复杂度帧内帧间压缩算法研究
发布时间：2019-06-29 14:02
【摘要】：当今信息化的时代,随着计算机及互联网技术的发展,视频应用越来越广泛而频繁。无线视频传输是无线射频收发和视频信息处理技术的结合,适用于许多难以靠有线传输视频的场合。当今各领域各应用场合对无线视频传输提出了更高的需求,视频分辨率和帧率的增大直接导致压缩后的视频码率骤然上升。为实现视频的及时传送,在提升无线收发能力的同时,采用新一代视频压缩标准HEVC能够节约一半的码率。为在无线视频传输系统中应用HEVC,需要解决其计算复杂度高的问题。目前的基于HEVC的低复杂度帧内帧间算法,一般针对的环节有限,虽然保证了率失真性能几乎不变,但编码时间的降低也比较有限。考虑到无线视频传输的应用实际,本文在略微牺牲率失真性能的情况下,节约了更多的编码时间。基于对帧内帧间视频压缩原理的理解和对最优率失真算法流程的研究,本文对帧内帧间的CU划分环节和模式预测环节都进行了低复杂度改进。本文首先提出了基于图像空间相关性的CU快速划分算法,利用相邻CU的深度信息为当前CU设定深度范围,该算法可以同时应用于帧内预测与帧间预测的情形。对于帧内预测,本文首先根据参考像素方差和帧内模式之间的统计规律,提出了可以跳过RMD过程并排除角度模式的快速算法。在RMD与RDO模式选择过程中,本文利用图像纹理特征以及MPM统计特征实现了帧内模式的快速优选。接着利用帧内最终预测模式的RD代价实现了CU划分的提前终止,最后结合上述各改进点设计了低复杂度帧内预测综合算法。对于帧间预测,本文基于对率失真代价与预测残差及量化参数的理解,认为当满足Skip条件或CBF为零的条件时,为实现更高的编码时间节约比率,可以不加约束地跳过后续其他模式。另外,考虑到非对称分割模式(AMP)起到的作用较小,本文禁用了非对称模式。最后结合上述帧间预测策略及CU快速划分算法,设计了低复杂度帧间预测综合算法。实验表明,相对于现有低复杂度算法,本文提出的帧内帧间综合算法,在率失真性能损失略微提升的前提下,可以降低更多的编码时间。
[Abstract]:Nowadays, with the development of computer and Internet technology, video application is more and more extensive and frequent. Wireless video transmission is a combination of radio frequency transceiver and video information processing technology, which is suitable for many occasions where it is difficult to transmit video by cable. Nowadays, there is a higher demand for wireless video transmission in various fields. The increase of video resolution and frame rate directly leads to the sudden increase of video rate after compression. In order to realize the timely transmission of video, the new generation video compression standard HEVC can save half the bit rate while improving the wireless transceiver ability. In order to apply HEVC, in wireless video transmission system, it is necessary to solve the problem of high computational complexity. The current low complexity intra-frame algorithm based on HEVC is generally limited. Although the rate-distortion performance is almost unchanged, the reduction of coding time is also limited. Considering the application practice of wireless video transmission, this paper saves more coding time under the condition of slightly sacrificing rate-distortion performance. Based on the understanding of the principle of intra-frame video compression and the research of the optimal rate-distortion algorithm, this paper improves the CU partition and mode prediction of intra-frame frames with low complexity. In this paper, a fast CU partition algorithm based on image spatial correlation is proposed. Using the depth information of adjacent CU to set the depth range for the current CU, the algorithm can be applied to both intra-frame prediction and inter-frame prediction. For intra prediction, according to the statistical law between reference pixel variance and intra mode, a fast algorithm which can skip RMD process and exclude angle mode is proposed in this paper. In the process of RMD and RDO mode selection, this paper uses image texture features and MPM statistical features to achieve fast intra mode selection. Then the early termination of CU partition is realized by using the RD cost of the final intra prediction mode. Finally, a low complexity intra prediction synthesis algorithm is designed according to the above improved points. For inter-frame prediction, based on the understanding of rate-distortion cost, prediction residual and quantification parameters, it is considered that when the Skip condition or CBF is zero, in order to achieve a higher coding time saving ratio, the following other modes can be skipped without constraint. In addition, considering that asymmetric segmentation mode (AMP) plays a small role, asymmetric mode is disabled in this paper. Finally, combined with the above inter-frame prediction strategy and CU fast partition algorithm, a low complexity inter-frame prediction synthesis algorithm is designed. The experimental results show that compared with the existing low complexity algorithms, the intra-frame synthesis algorithm proposed in this paper can reduce the coding time with the slight improvement of rate-distortion performance loss.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81
本文编号：2507868
[Abstract]:Nowadays, with the development of computer and Internet technology, video application is more and more extensive and frequent. Wireless video transmission is a combination of radio frequency transceiver and video information processing technology, which is suitable for many occasions where it is difficult to transmit video by cable. Nowadays, there is a higher demand for wireless video transmission in various fields. The increase of video resolution and frame rate directly leads to the sudden increase of video rate after compression. In order to realize the timely transmission of video, the new generation video compression standard HEVC can save half the bit rate while improving the wireless transceiver ability. In order to apply HEVC, in wireless video transmission system, it is necessary to solve the problem of high computational complexity. The current low complexity intra-frame algorithm based on HEVC is generally limited. Although the rate-distortion performance is almost unchanged, the reduction of coding time is also limited. Considering the application practice of wireless video transmission, this paper saves more coding time under the condition of slightly sacrificing rate-distortion performance. Based on the understanding of the principle of intra-frame video compression and the research of the optimal rate-distortion algorithm, this paper improves the CU partition and mode prediction of intra-frame frames with low complexity. In this paper, a fast CU partition algorithm based on image spatial correlation is proposed. Using the depth information of adjacent CU to set the depth range for the current CU, the algorithm can be applied to both intra-frame prediction and inter-frame prediction. For intra prediction, according to the statistical law between reference pixel variance and intra mode, a fast algorithm which can skip RMD process and exclude angle mode is proposed in this paper. In the process of RMD and RDO mode selection, this paper uses image texture features and MPM statistical features to achieve fast intra mode selection. Then the early termination of CU partition is realized by using the RD cost of the final intra prediction mode. Finally, a low complexity intra prediction synthesis algorithm is designed according to the above improved points. For inter-frame prediction, based on the understanding of rate-distortion cost, prediction residual and quantification parameters, it is considered that when the Skip condition or CBF is zero, in order to achieve a higher coding time saving ratio, the following other modes can be skipped without constraint. In addition, considering that asymmetric segmentation mode (AMP) plays a small role, asymmetric mode is disabled in this paper. Finally, combined with the above inter-frame prediction strategy and CU fast partition algorithm, a low complexity inter-frame prediction synthesis algorithm is designed. The experimental results show that compared with the existing low complexity algorithms, the intra-frame synthesis algorithm proposed in this paper can reduce the coding time with the slight improvement of rate-distortion performance loss.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81
【参考文献】
相关期刊论文 前1条
1 郭奕星;;下一代数字视频技术应用前瞻[J];电信快报;2014年08期
相关博士学位论文 前1条
1 王建富;H.265/HEVC编码加速算法研究[D];中国科学技术大学;2015年
相关硕士学位论文 前2条
1 蒋亮;基于ARM的无线图像传输系统设计[D];电子科技大学;2013年
2 安中文;一种无线图像传输系统设计方案[D];北京邮电大学;2013年
，本文编号：2507868
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2507868.html
