基于感兴趣区域的HEVC优化算法研究
发布时间：2019-05-09 07:52
【摘要】：随着视频分析、处理和传输技术要求的不断提高,视频监控系统已由原来的模拟方式向数字化、智能化方向改进,但却导致了视频监控数据的急剧增加,使得现有的视频编码标准已不能满足新型视频应用的发展要求。新一代高效视频编码标准H.265/HEVC(High Efficiency Vedio Coding)提供了多项新的技术,随着自身不断完善的同时将逐渐取代H.264/AVC,并逐步应用于各类视频编码业务中。HEVC采用的新技术使得编码效率相比于H.264提高了一倍,但这些新技术也使得编码计算复杂度成倍增加。近年来,如何采用有效的方法减小HEVC的编码计算复杂度一直是HEVC视频编码研究工作的重中之重。为了提高编码效率,本文将智能视频监控技术和HEVC编码标准相结合,从HEVC编码框架的关键模块入手,深入学习各关键编码模块的技术原理,通过详细研究编码单元块(Coding Unit,CU)的划分算法和熵编码技术原理,提出了基于高关注度区域对CU块划分和熵编码技术的改进算法。首先为本文的改进算法做前期工作,提取视频序列中的高关注区域,也称作感兴趣区域即监控视频中的运动目标区域,感兴趣区域主要利用Kinect和立体视频的深度信息和混合高斯模型法提取视频序列中的背景信息结合深度信息定义和背景差分法得到;其次将提取的感兴趣区域关联到视频的编解码过程中,针对已获取的感兴趣区域,对HEVC中CU块划分提出提前终止策略,减少不必要的率失真代价计算,加快编码效率,该方法在保证视频编码质量损失较小的前提下,平均编码时长缩减54.35%,显著提升了编码效率;最后考虑到CU块划分对熵编码的影响,利用感兴趣区域对原始率失真算法进行改进,在保证视频图像质量和编码复杂度相对不变的情况下,减少了51.6%的比特率,明显改善了HEVC原有算法的压缩性能。
[Abstract]:With the continuous improvement of video analysis, processing and transmission technology, video surveillance system has been improved from the original analog mode to the direction of digitalization and intelligence, but it has led to a sharp increase of video surveillance data. So that the existing video coding standards can no longer meet the requirements of the development of new video applications. The new generation of efficient video coding standard H.265/HEVC (High Efficiency Vedio Coding) offers a number of new technologies that will gradually replace H.264 AVC as it continues to be perfected. The new technologies adopted by HEVC make the coding efficiency more than double that of H.264, but these new technologies also double the complexity of coding. In recent years, how to use effective methods to reduce the computational complexity of HEVC coding has been the most important task in the research of HEVC video coding. In order to improve the coding efficiency, this paper combines the intelligent video surveillance technology with the HEVC coding standard, starts with the key modules of the HEVC coding framework, deeply studies the technical principles of the key coding modules, and studies the coding unit block (Coding Unit, in detail. Based on the partition algorithm of CU and the principle of entropy coding, an improved algorithm for block partition and entropy coding of CU based on high attention region is proposed. First of all, we do the preliminary work for the improved algorithm in this paper, and extract the high attention region in the video sequence, which is also called the region of interest, that is, the moving target area in the surveillance video. The region of interest mainly uses the depth information of Kinect and stereo video and the mixed Gao Si model method to extract the background information from the video sequence combined with the definition of depth information and the background difference method. Secondly, the extracted region of interest is associated with the video encoding and decoding process. Aiming at the acquired region of interest, an early termination strategy is proposed for the partition of CU blocks in HEVC, which reduces unnecessary rate-distortion cost calculation and speeds up coding efficiency. On the premise that the video coding quality loss is small, the average coding time is reduced by 54.35%, and the coding efficiency is significantly improved. Finally, considering the effect of CU block partition on entropy coding, the original rate-distortion algorithm is improved by using the region of interest, and the bit rate is reduced by 51.6% while the video image quality and coding complexity are relatively unchanged. The compression performance of the original HEVC algorithm is improved obviously.
【学位授予单位】：中原工学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81
本文编号：2472600
[Abstract]:With the continuous improvement of video analysis, processing and transmission technology, video surveillance system has been improved from the original analog mode to the direction of digitalization and intelligence, but it has led to a sharp increase of video surveillance data. So that the existing video coding standards can no longer meet the requirements of the development of new video applications. The new generation of efficient video coding standard H.265/HEVC (High Efficiency Vedio Coding) offers a number of new technologies that will gradually replace H.264 AVC as it continues to be perfected. The new technologies adopted by HEVC make the coding efficiency more than double that of H.264, but these new technologies also double the complexity of coding. In recent years, how to use effective methods to reduce the computational complexity of HEVC coding has been the most important task in the research of HEVC video coding. In order to improve the coding efficiency, this paper combines the intelligent video surveillance technology with the HEVC coding standard, starts with the key modules of the HEVC coding framework, deeply studies the technical principles of the key coding modules, and studies the coding unit block (Coding Unit, in detail. Based on the partition algorithm of CU and the principle of entropy coding, an improved algorithm for block partition and entropy coding of CU based on high attention region is proposed. First of all, we do the preliminary work for the improved algorithm in this paper, and extract the high attention region in the video sequence, which is also called the region of interest, that is, the moving target area in the surveillance video. The region of interest mainly uses the depth information of Kinect and stereo video and the mixed Gao Si model method to extract the background information from the video sequence combined with the definition of depth information and the background difference method. Secondly, the extracted region of interest is associated with the video encoding and decoding process. Aiming at the acquired region of interest, an early termination strategy is proposed for the partition of CU blocks in HEVC, which reduces unnecessary rate-distortion cost calculation and speeds up coding efficiency. On the premise that the video coding quality loss is small, the average coding time is reduced by 54.35%, and the coding efficiency is significantly improved. Finally, considering the effect of CU block partition on entropy coding, the original rate-distortion algorithm is improved by using the region of interest, and the bit rate is reduced by 51.6% while the video image quality and coding complexity are relatively unchanged. The compression performance of the original HEVC algorithm is improved obviously.
【学位授予单位】：中原工学院
【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10条
1 詹晓涛;;HEVC帧内预测编码快速判决算法分析[J];中国有线电视;2015年07期
2 王璐;;智能视频监控技术应用[J];中国科技信息;2015年05期
3 黄凯奇;陈晓棠;康运锋;谭铁牛;;智能视频监控技术综述[J];计算机学报;2015年06期
4 许东旭;林其伟;;高效率视频编码帧内快速算法[J];计算机应用;2014年08期
5 于成业;何小海;滕奇志;钟国韵;李元;;基于非规则PU模式的HEVC视频编码方法[J];电视技术;2014年09期
6 阮若林;胡瑞敏;;数字视频压缩编码技术标准现状与展望[J];电视技术;2014年03期
7 战荫伟;张昊;;基于Kinect传感器的人体行为分析算法[J];传感器与微系统;2015年01期
8 钟国韵;汤彬;徐洪珍;吴有用;;时空域相关性的快速HEVC帧间模式判决方法[J];东华理工大学学报(自然科学版);2013年04期
9 徐飞;张永飞;李哲;;HEVC熵编码算法优化[J];计算机与现代化;2013年12期
10 沈燕飞;李锦涛;朱珍民;张勇东;;高效视频编码[J];计算机学报;2013年11期
，本文编号：2472600
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2472600.html
