基于人类视觉特性的结构相似度图像质量评价

【作者】 王经谊；

【导师】 张毅；

【作者基本信息】 南京理工大学 ， 光学工程， 2012， 硕士

【摘要】 随着网络及数字技术的发展,数字图像已广泛深入人们的日常生活,成为信息传递的重要载体。然而在现实生活中,由于各种因素的影响会使图像产生失真,影响传递的信息的质量,因此开展图像质量评价技术的研究,具有十分重要的现实意义。主观评价方法是最准确的评价方法,但是耗时耗力,并且过程十分繁琐,因此在实际中很难应用。均方差和峰值信噪比等传统的客观评价方法不符合人的主观感受。一些简单地模拟人眼功能的图像质量评价方法能够较好地反映人的主观感受,但是人类视觉特性十分复杂,且目前的研究尚不够深入,所以其发展受到很大的限制。Zhou Wang等人提出了一种基于结构相似度的图像质量评价方法,该方法简单高效,在主客观一致性方面取得了更好的效果。本文在深入研究人类视觉特性和结构性相似理论等地基础上,提出一种基于人类视觉特性的结构相似度图像质量评价方法。通过小波变换将原始图像和待测图像分解为不同频带和方向上的子带图像,分别计算各个对应子带图像的结构相似度值,并利用对比度敏感函数曲线计算人眼对每幅子带图像的敏感度,最终将所有图像的相似度值加权处理得到一个综合的测度值。通过仿真分析,结果表明本文提出的方法比传统的客观评价方法和基于结构相似度的评价方法更有效,其客观评价结果与人的主观感受具有更高的一致性。更多还原

【Abstract】 With the development of the internet and digital technology, digital image has gone deep into people’s daily lives thoroughly, which has become the most important carrier of information. However, the image will always be distorted by varities of factors in real life, which will reduce the quality of the information. Therefore, the study of image quality assessment technique has a very important practical significance.Subjective evaluation is the most accurate evaluation method, but it costs too much manpower and time, and the process is too complicated to use in practice. Some traditional objective evaluation methods, such as MSE and PSNR, cann’t reflect people’s subjective feelings. Some methods of image quality evaluation which can simpy simulate the function of human eyes can reflect the people’s subjective feelings better, but they are limited by the complexity of HVS and the unsolved problems. Zhou Wang proposed a simple and efficient method which based on the structure similarity theory, this method, too some extent, can objectively reflect people’s subjective feelings.This paper through in-depth research on HVS and structure similarity theory to propose a structure similarity method based on HVS. Firstly, the original image and the target image are decomposed into sub-band images in different frequency and directions by wavelet transform, and the SSIM values of every sub-band image are calculated according to the SSIM method. And then, the sensitivity values of each sub-band image will be got by using the contrast sensitivity function curve. Finally, a comprehensive value can be obtained by weighting the SSIM values of all sub-band images. Through the simulation and analysis, the results show that the method proposed in this paper is more effective than the traditional evaluation method and the SSIM method, and it reflects the people’s subjective feeling with higher accuracy.更多还原