【作者】 姚志恒；

【导师】 沈兰荪；

【作者基本信息】 北京工业大学 ， 信号与信息处理， 2002， 硕士

【摘要】 通过无线和互联网传输视频是视频通信的发展趋势。然而由于有限的带宽和恶劣的通信环境，使得窄带高噪声成为无线和互联网视频通信的难点。本文研究了基于H.263视频压缩编码标准的视频传输若干关键问题，包括：视频传输的抗误码技术、快速高效的压缩编码技术、视频质量评价方法等，并且实现了一个局域网内视频传输演示实验系统。 如何消除或减小噪声对视频信息的影响是视频传输中的重要问题。前向纠错的方法可以提高压缩后的视频码流在传输中抵抗误码的能力，但要以降低压缩效率为代价。由于视频信息具有非常强的时间空间相关性，因而可以通过误码掩盖的方法减小误码对视频质量造成的影响。本文针对因特网视频传输中的丢包现象进行研究，一方面利用RS编码来提高视频压缩码流抵抗误码的能力；另一方面，提出了一种基于运动特征的自适应误码掩盖方法，能够较好的掩盖视频传输带来的误码，特别对于运动幅度较大的视频序列更能体现出其优势。 数据量大是视频信息的突出特点，较小的时延抖动是视频通信最基本的要求，因此高效快速的视频压缩技术成为视频传输的关键。本文以H.263压缩编码标准为基本框架，针对其中影响压缩效率和速度的瓶颈展开研究，包括快速运动搜索算法的研究、预判零方法的研究以及MMX技术在视频编码中的应用等，提出了一种基于时域预测的预判零新方法，在不影响重建视频质量的条件下具有较高的判别效率，节约了大量DCT变换和量化的时间，提高了编码的速度，达到了实时性要求。 视频质量是衡量视频传输服务质量的重要指标，而对其的评价是一个与人眼视觉特性密切相关的过程。本文提出一种基于运动特征的视频质量评价方法，考虑了人眼对视频信号感受质量与画面运动剧烈程度之间的关系，分别讨论了反映视频质量的视频清晰度和流畅度，评价结果更符合人眼视觉特性，并且这种评价对编码具有一定的指导作用。 综合以上研究，实现了一个局域网内视频传输演示实验系统，该系统包括采集、压缩编码、传输、解码和显示等环节，从而验证了研究成果的实用性和有效性。更多还原

【Abstract】 Video transmission over unreliable channels such as wireless channels and Internet has become trend of video communication. However, there are some inherent problems in wireless channels and Internet that affect video delivery, such as limited bandvvidth, varying delay and high levels of noise. Some key issues in video transmission based on H.263, including video compression, error resilient and video quality evaluation are studied in this paper. Finally, a video transmission DEMO in LAN is implemented.When a video stream is conveyed through a vvireless channel or Internet to the receiver end, it is exposed to high levels of bit errors or packet loss. These errors severely degrade the service quality of video communication due to the propagation and accumulation of errors throughout the transmitted video stream. There are several procedures used to reduce these deteriorating effects, including the use of Forward Error Correction techniques, error concealment methods and error resilience schemes.This paper studies error resilient techniques in Internet video transmission. In this techeniques, Reed-Solomon(RS) codes across packets is used for error protection against packet loss, and a self-adaptive error concealment method based on motion feature is proposed, which can conceal errors more effectively, particularly for video sequences with acute motion.Due to limited bandvvidth and stringent requirements of real-time video plavback, video coding is an indispensable process for many visual communication applications and always requires a very high compression ratio. Generally speaking, a coding algorithm must be low complexity and work at high speed. In this paper, the coding methods are studied to increase coding speed. Firstly, a novel method based on temporal prediction to detect all-zero DCT coefficients is proposed, which can greatly reduce the coding cost and maintain a good tradeoff between correct detecting ratio and false detecting ratio with little additional computation. Secondly, a new diamond search (DS) algorithm for fast block-matching motion estimation is implemented. We also map some of the instructions onto IntePs MMX architecture. Using the above algorithms, H.263 baseline video-encoder implementation can encode more than 40fps in QCIF resolution on a Pentium MMX 500-MHz computer.Almost ali of the low bit-rate video compression techniques degrade quality. To combine the video quality evaluation with HVS, the relationship betvveen the motion activity and perceptual video quality are studied, and a video quality evaluation method based on motion activity is proposed, in which video defmition and smoothness are defined.Finally, a near real-time software-only DEMO is presented to demonstrate the overall procedure from video acquirement, video coding, transmission, to video decoding and playing, which confirm the practicability and efficiency of above techeniques.更多还原