一种新的基于局部特征的图像质量评价方法

传统的基于结构相似度（SSIM）的质量评价方法具有适用范围狭窄，评价算法不稳定等特点。在对传统图像质量评价算法研究的基础上，提出了一种新的基于局部特征的质量评价方法。与传统方法不同，在对图像质量进行评价时，该方法充分考虑到图像的结构信息对于图像质量的影响。新的方法主要分为3个步骤：首先，基于一种新的图像分块算法，根据图像的结构信息将图像划分成不同的块；其次，利用图像的梯度作为衡量像素重要程度的权值，计算参考图像和失真图像对应图像块的结构相似度；最后，融合各个块的相似度信息获得最终的图像质量评价结果。实验结果表明，该方法的评价结果更加合理、稳定，适用范围广，优于传统的基于结构相似度的质量评价方法。

A Novel Image Quality Assessment Method Base on Local Character

Image quality is mainly affected by its structure and content. Traditional image quality evaluation metrics based on structural similarity put emphasis on image structure, but inadequately consider local features of image. So their application fields are limited and performances are unstable. If dividing an image into more meaningful structural blocks, the impaction of local features on image quality can be represented adequately and metric performance can be improved greatly. Based on these considerations, this paper proposes a new quality index using local character. It is implemented by three steps. Firstly, the image is divided into separate meaningful blocks according to a new image division algorithm. Different blocks represent different structures of the image. Secondly, the gradient of the image is used to weigh the influence of different pixels, and then the structural similarities of corresponding blocks between the reference image and distorted image are calculated. Finally, the ultimate image quality is calculated by combining structural similarities of all blocks according to their weights. The experiments show that the proposed metric is more reasonable and stable than traditional methods, and could be used in more application fields.