峰值信噪比（PSNR）是常用的衡量信号失真的指标，但是PSNR不涉及信号自身内容的特征，对某些图像或视频序列进行质量评价时会与主观感知的质量产生较大的偏差。结构相似法SSIM（structural similarity）是一种基于结构信息衡量原始信号与处理后信号之间相似程度的方法，计算简单、与主观质量评价关联性较强。为此提出将PSNR和SSIM联合起来建立图像质量评价模型，先利用聚类分析法根据PSNR值和SSIM输出值对样本图像进行规整聚类，然后对不同类别的图像运用不同的质量评价规则，评价规则由二元回归分析确定；待测图像通过支持向量机（support vector machines，SVM）分类器实现分类。实验结果表明，该模型的输出能有效地反映图像的主观质量。

PSNR(peak signal noise ratio) is the common criterion used to assess the distortion of signal.But big error maybe generated with PSNR which does not involve the content of signal when used to assess image quality.SSIM(structural similarity) is used to evaluate the similarity between the source signal and the processed signal.SSIM is simple and well correlated with subject evaluation.This paper,PSNR and SSIM are combined to set up the image quality assessing model.Cluster analysis is used to make the samples data cluster into different kinds.Support Vector Machines Classifier is used to class any image into different kinds according to PSNR and SSIM.The quality of the image with different kinds is assessed with different strategy.The results from our test show the model output can reflect the image subjective quality effectively.