随着计算机软硬件技术的发展，数字图像应用越来越广泛。然而，在图像采集、处理、传输和呈现的过程中，都可能引入各种失真。主观方法和传统客观方法由于自身的局限性，不能达到应用需求。因此，研究与人眼主观视觉感知相一致的客观图像质量评价算法具有重要意义。本文深入分析了结构相似性理论，比较分析了传统方法、绝对误差和SSIM在图像质量评价中的性能和表现，并且结合SSIM，从图像内容、图像划分、颜色相似性方面对算法做出改进。 针对质量评价未考虑人眼主观感知的问题，本文模拟人眼观察图像时，会不自觉的按照图像边界将图像内容划分为多个区域的特性，提出了基于图像内容划分的结构相似度质量评价算法(Partition SSIM, PSSIM)。该方法结合参考图像和待测图像，使用图像梯度，将图像划分为变化边界区域、不变边界区域、纹理区域和平坦区域，对于图像各个区域，通过训练方法得到权重参数，最后融合为单一质量描述符。实验表明，在SSIM值相同而主观感受不一致的图像中，PSSIM有更好的准确性。在LIVE图像数据库上的实验证明PSSIM方法比PSNR 、SSIM、SS_SSIM等方法与主观感知有更好的一致性。 人眼对于颜色敏感，在彩色图像失真中，颜色失真占有很大比例，颜色失真几乎与结构失真一样严重影响图像质量。原始的SSIM方法在灰度化过程中忽略了颜色信息，度量颜色失真可以大大提升客观图像质量评价的准确性。本文首先提出了一种颜色相似度测量方法，将颜色空间RGB转换到均匀的HIS空间，使用指数函数来模拟颜色视觉降质。针对真彩色图像质量评价，本文结合颜色相似性和结构相似性，提出了一种基于颜色和结构双重相似度的质量评价方法(Color Similarity SSIM，CS_SSIM)，该方法同时度量图像颜色失真和结构失真，基于图像颜色信息对图像进行区域划分，从而得到图像的客观质量评价。在真彩色图像的测试实验中，在颜色失真而结构未受到严重破坏的图像中，该方法可以较好的度量图像质量的主观评价差异。

As the development of computer technology on hardware and software, digital image applications are used more and more widely. However, the images can introduce all kinds of distortion in the process of image acquisition, processing, transmission and rendering. Due to the limitations of its own, subjective methods and traditional objective methods can not reach the requirement of applications. So, it is significant to research an objective image quality assessment method that consistent with human subjective perception. In this paper, we deeply analysis the structure similarity theory, revise the calculation of correlation and compare the performance of traditional method, the absolute error method and SSIM on image quality assessment. We proposed some improvements of SSIM based on image content, image partition and image color similarity. In order to overcome the weakness of image quality assessment which not consider the human subjective perception. This paper simulates the human perception feature that the images will unconsciously be divided into multiple areas when human beings observe. And these areas generally are divided according to boundary. Simulation of the feature, this paper presents a structure similarity quality evaluation algorithm based on image content (Partition SSIM, PSSIM). This method uses the gradient of reference image and test image to divide the image into four areas, the preserved edge area, changed edge area, texture area and smooth area. For each area, we obtain the weighting parameters by training method. Finally, use the SSIM map and weighting parameters, we merge all the data into a single quality descriptor. The experiments on images with equal SSIM value but have different perception shows that the PSSIM has a better accuracy. The experiment on LIVE image database shows that PSSIM has a better consistency with subjective perception than PSNR, SSIM, SSI_SSIM and other methods. Our eyes are sensitive to colors. Color distortion occupies large proportion in real-color images distortion. For real-color images, color distortion is as important as structure distortion. The SSIM ignores color information in graying images. The accuracy of image quality assessment will greatly improve if we measure the color distortion. In this paper, we proposed a new method that converts the color space from RGB to HSI space to erase the nonuniformity of RGB space and uses exponential function to simulate the difference of color to evaluate the color similarity. Combining color similarity and structure similarity, we proposed a new approach called CS_SSIM (Color Similarity SSIM) to evaluate quality of real color images. This method measures both color and structure distortion. Then use color information to divide the images into four areas. Finally, the quality assessment value is obtained by all the calculated data. Our experiment on the real-color images shows that this method can excellently evaluate the decrease of images quality if images have heavily color distortion but less structure distortion.