【作者】 李成文；

【导师】 宋建新；

【作者基本信息】 南京邮电大学 ， 信号与信息处理， 2011， 硕士

【摘要】 随着多媒体通信技术的发展,数字图像以及视频的应用深入到了人类生活的各个方面。数字图像视频在采集、处理、压缩、传输以及重建的过程中可能会引入各种各样的失真,因此经常需要测量数字图像视频的质量。在实际应用中,主观质量评价方法通常不方便,并且费时费力;而传统的客观质量评价方法,如MSE和PSNR,在很多情况下不符合人的主观感受。因此,有必要研究更符合人的主观感受的客观质量评价方法。结构相似性理论认为人眼视觉系统(HVS)的主要功能是从视觉场景中提取结构信息,因此对结构信息变化的度量可以作为图像感知失真的很好的近似。本文深入研究了结构相似理论的实现结构相似度(SSIM)指数在图像和视频评价方面的应用,改进了基于结构相似性的视频质量评价方法并将结构相似性应用于视频通信。一般来说,运动表征的质量将决定视频的感知质量,然而现有的大多数视频质量评价方法很少直接用到运动信息,因而限制了其有效性。另外,HVS对观察到的视频的质量较差的区域比较敏感。为了弥补这些缺陷,改进一种基于结构相似性的空时视频质量评价方法(stVSSIM),其分别在时域和空域上对视频的质量进行估计,并根据HVS的感知特性对质量指数进行相应的处理。实验结果表明,stVSSIM方法比MOVIE方法以及PSNR等方法的有效性和准确性更高。在视频通信方面,本文改进一种基于结构相似性的H.264快速运动估计算法( FMESS)。该算法应用SSIM值较大的两个图像块之间的差信号通常是一个低频信号,而低频信号通常比较容易压缩的特点以及SSIM值是图像块的结构信息的反映,非常适合于提前结束算法的特性。通过在运动估计过程中设置固定的SSIM阈值,减少不必要的运动搜索,从而达到减小运动估计复杂度、改善压缩性能的目的。实验结果表明,本文所改进的FMESS算法在保证视频的压缩质量的前提下,不仅降低了视频编码的码率,而且节省了编码时间。更多还原

【Abstract】 Recent advances in multimedia communication technology have resulted in the proliferation of digital images, both still and video. However, digital image and video are subject to a wide variety of distortions during acquisition, processing, compression, storage, transmission and reproduction, so it’s always need to measurement its quality. In practice, however, subjective evaluation is usually too inconvenient, time-consuming and expensive. The traditional objective methods such as MSE and PSNR have low correlation with the perceptual visual quality. So it’s necessary to develop new objective methods which can correspond better to subjective feelings.The philosophy of the structural similarity based on the assumption that the human visual system is highly adapted to extract structural information from the viewing field. It follows that a measure of structural information change can provide a good approximation to perceived image distortion. This paper studies the applications of the Structural SIMilarity(SSIM)index on image and video quality assessment, and develop its applications on the video communication.Naturally, the quality of motion representation in videos plays an important role in the perception of video quality, yet existing VQA algorithms make little direct use of motion information, thus limiting their effectiveness. On the other hand, the human visual system is very sensitive to the low quality of the video detected. We seek to ameliorate this by developing a new VQA algorithm based on the structural similarity. Video quality is evaluated not only in space, but also in time. It is found that the stVSSIM index delivers VQA scores that correlate quite closely with human subjective judgment.On the other hand, we propose a novel H.264 fast motion estimation algorithm based on the structural similarity, the characteristics of SSIM are considered, a SSIM threshold is set for the motion estimation process and the unnecessary searching positions are eliminated, thus reducing the complexity of motion estimation and improving the coding performance. It is found that the FMESS effectively saves the coding time and improves the compression ratio without reducing the video quality.更多还原