基于视觉显著性和稀疏表示的图像质量评价

项目名称： 基于视觉显著性和稀疏表示的图像质量评价

项目编号： No.61201394

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 张林

作者单位： 同济大学

项目金额： 24万元

中文摘要： 如今，图像处理技术的应用已渗透到国民经济的各个行业中。在几乎所有的图像应用系统中，图像的质量评价都扮演着基石的角色。图像质量评价研究的最终目的是希望提出某些算法，这些算法可以对图像的质量进行自动评价，而且评价的结果能够和人的主观感受高度一致。目前的图像质量评价方法（尤其是无参考图像质量评价方法）对图像质量的评价能力还达不到和人的主观评测高度一致的程度。鉴于此，在本项目中，我们将致力于提出新的高性能的图像质量评价方法。拟提出的新方法将建立在视觉显著性和稀疏表示两个基础理论之上。这是由于：一方面，图像的视觉显著性特征与人对图像的主观感知之间有着密切的联系；另一方面，稀疏表示理论是近年来发展起来的有效的图像表示和分类模型。我们将尝试把这两个理论有机结合，并有效地引入到图像质量评价领域中，来构建全新的高性能的图像质量评价方法。本项目的预期成果将在理论和应用两个方面极大地促进图像处理领域的发展。

中文关键词： 图像质量评价；视觉显著性；稀疏表示；自然场景统计模型；

英文摘要： So far, kinds of image processing systems have been used in various applications. Image quality assessment (IQA) plays a significant and fundamental role in almost all of these systems. Actually, the ultimate goal of the IQA research is to design some computational models that can automatically measure the image quality in consistency with subjective ratings. Currently, however, the quality scores predicted by the modern IQA indices, especially the no reference IQA indices, cannot highly correlate with subjective evaluations. To this end, we will lay our focus on devising novel high performance IQA indices in this project. Specifically, the newly proposed IQA indices will be based on computational visual saliency models and the sparse representation theory, for which there are several justifications. On one hand, an image's visual saliency map has a close relationship to its perceptual visual quality. On the other hand, the sparse representation theory developed in recent years is an effective tool for image representation and classification and it can be made use of to discriminate high-quality image patches from low-quality ones. Thus, in this project, we will attempt to integrate these two theories together appropriately and to introduce them into the IQA community. And accordingly, based on them, we will pro

英文关键词： image quality assessment；visual saliency；sparse representation；natural scene statistics model；