Model-based full reference image blurriness assessment

详细信息    查看全文

• 作者：Mohammad Hossein Khosravi ; Hamid Hassanpour
• 关键词：Image blurriness evaluation ; Structural information ; Low frequency redundancies ; Singular value decomposition
• 刊名：Multimedia Tools and Applications
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：76
• 期：2
• 页码：2733-2747
• 全文大小：
• 刊物类别：Computer Science
• 刊物主题：Multimedia Information Systems; Computer Communication Networks; Data Structures, Cryptology and Information Theory; Special Purpose and Application-Based Systems;
• 出版者：Springer US
• ISSN：1573-7721
• 卷排序：76

文摘

Measuring image blurriness is an important issue in image-quality assessment. The blurriness affects the image quality by degrading the image’s high frequency details in the form of some uniform redundancies in neighboring pixels. Indeed, the blurriness is accompanied with two destructions: corrupted high frequency details, and degraded image structure due to the redundancies. In this paper, we propose an approach that measures the effects of these two distortions using singular value decomposition (SVD). From the properties of SVD, the basis images corresponding to the higher singular values are associated with the structural information of the image, while the ones corresponding to the lower singular values are related to the image details. This work employs this property and splits the ordered singular values into two subsets from a non-fixed separation point, and constructs two images by stacking up the basis images corresponding to these two subsets. By moving the separation point for these two subsets and computing the energy of the two constructed images in each point, two sequences of energies will be in hand. We shows that the behavior of these two sequences can be used to assess the amount of both structural distortions and nonstructural detail degradations of an image, and hence a valuable blur metric. Experimental results illustrate that there is a well correlation between the results of our blur metric and human scores. In addition, in comparative experiments, we found that the proposed blur metric is stand among the best state-of-the-art ones in evaluating quality of images in terms of blurriness.