An Efficient No-Reference Blurriness Metric for Images and Video frames

The objective measurement of blurring artifacts plays an important role in the dynamic monitoring, adjusting image quality, optimizing algorithms and parameter settings in a variety of image and video processing applications. In this paper, we propose a no-reference metric to estimate blurriness distortion. The work addresses the problem that existing no reference metrics fail to predict the correct amount of blurriness in images and video with varying content. The proposed metric uses a human vision system (HVS) based approach to detect the edges in the images and also uses singular value decomposition (SVD) to estimate the blurriness. The metric performance is demonstrated by comparing with other existing no-reference metrics for two different publicly available data bases. Test results show that the proposed metric exhibits a strong correlation with subjective quality scores.