视频编码和编码转换中的运动矢量估计

【作者】 董武；

【导师】 李晓辉；

【作者基本信息】 安徽大学 ， 电路与系统， 2004， 硕士

【摘要】 随着无线通信的广泛应用，视频图像的编码和传输技术面临巨大的挑战。由于视频的数据量巨大，为了满足视频在频带受限的无线信道上传输的实时性要求，必须使用数据量压缩比大而且复杂度小的快速编码算法，尽量用最少的数据传输最大的信息量。运动估计是运动图像压缩中的关键技术之一，视频信号在时间上有很强的相关性，利用块匹配估计和运动补偿技术，可以有效地去除图像帧间冗余度，实现高压缩比。通常，在编码器运行中，运动估计算法需要消耗70％左右的执行时间，因此为了提高编码器的速度必须首先提高运动估计算法的效率。 此外，由于Internt和移动通信的高速发展，出现了各种具有不同性能的客户机，如蜂窝手机、PDA、手提电脑和膝上电脑等等，这些客户机迫切要求能够无线接入Internet，浏览Internet上的内容。由于Internet和无线网络具有不同的带宽，因而也就对应着不同的传输码率。如果将己压缩的视频信号流由互联网直接通过无线网络传送给客户机，将会出现视频编码流与传输信道失配的情况。此时，就需要在Internet和客户机之间设置代理服务器，对已压缩编码的视频信号流进行码率转换，将已压缩的高速视频码流转换成低速率的视频码流，以保证视频信号流在移动无线网络中的正确传输，为移动用户提供不同服务质量的视频服务。 本文在以下几个方面进行了研究： (1)在传统的运动估计算法的基础上，根据视频中物体的运动情况提出了一种基于图像运动特征的快速运动估计算法。仿真实验结果表明，本文算法适用于各种运动情况的图像序列，其性能接近于全搜索算法，同时极大地降低了计算复杂度。 (2)介绍了视频编码转换中的各种转换模型。视频编码转换既可以在像素域中进行，也可以在变换域—DCT域中进行。具体的转换方式有三种：码率转换、分辨率转换、编码制式的转换。码率转换一般是降低视频的码率，提高不同网络的兼容性；分辨率转换一般是降低视频的空间分辨率和时间分辨率；编码制式的转换是对已用一种标准编码后的视频流用另外一种标准来编码。 (3)在基于降低时间分辨率的转换中，本文分析了合成运动矢量的线性内插法、FDVS方法，并在此基础上提出了一种运动矢量合成的新方法。仿真实验结果表明，该算法同已有的合成算法相比，不仅提高了视频的转换质量，而且提高了转换速度。更多还原

【Abstract】 With great applications of wireless communication, there are great challenges in the coding and transmission technology of video. Because data quantity of video is very big, fast coding algorithm used must have high data compression ratio and low complexity, and least data quantiy used can transmit most information quantity in order to satisfy video’s real time request, when video is transmitted in narrow wireless channel. Motion Estimation is one of key technology in the video image compression coding. Video signals have very high motion correlation in temporal direction, motion estimation and motion compensation technology based on block matching can eliminate redundancy of inter-frame’s effectively to achieve high compression ratio. Commonly, motion estimation algorithm consumes about 70% computing time in coder, so to improve the coder’s speed, the efficiency of motion estimation must be improved firstly.Furthermore, with Internet and mobile communication’s high development, various client devices having different performance appear, such as: cellular phone, PDA, hand-held computer, laptop computer, etc. These client devices hope to gain access to Internet in wireless channel, and browse Internet’s content. As Internet and wireless network have different bandwidth, they have different code rate accordingly. If video signal stream compressed is transmitted in wireless channel from Internet to client devices, coded video signal stream will not match the wireless channel. So, proxy need to be set between generic WWW servers and client devices, transcoding video signal stream which has been coded and compressed. Proxy can transform high speed video stream into low speed video stream to make video signal stream transmitted accurately in mobile wireless network, and provides client devices different quality of service.The works we have researched include several aspects as follows:(1) Traditional motion estimation algorithms are analysed, and a fast motion estimation algorithm based on video objects’ motion character is proposed. The simulation experiment result shows this algorithm is adapted to image sequences with different motion characteristic, and its performance is close to full search algorithm while greatly reducing computation complexity.(2 ) Various transcoding models in video transcoding are analysed. Video transcoding can be carried out not only in pixel domain but also in transform domain, namely DCT domain. Videotranscoding has three means: code rate conversion, resolution conversion, conversion between coding standard. Code rate conversion commonly is reducing video’s code rate. Resolution conversion commonly is reducing video’s temporal resolution and spatial resolution. Conversion coding standard is recoding video with another coding standard which has been coded with a standard.(3) In reducing temporal resolution transcoding two motion vector composed methods are analysed: linear interpolation method and FDVS method. A new motion vector composed method is proposed. Experiment results reveal the proposed algorithm is computationally efficient while keeping better image quality compared with the previous presented algorithms.更多还原