A No-Reference Metric for Evaluating the Quality of Motion Deblurring

• Yiming Liu¹
• Jue Wang²
• Sunghyun Cho²
• Adam Finkelstein¹
• Szymon Rusinkiewicz¹

• ¹Princeton University
• ²Adobe Research

ACM SIGGRAPH Asia 2013

We develop a no-reference metric for evaluating the perceptual quality of image motion deblurring results. The metric can be used for fusing multiple deblurring results (b) of the same input image (a) to generate one with the best quality (e). (c) and (d) are the result of simply averaging all deblurring results and the result of a naive fusion method, respectively. See text in Sec. 6.3 for more details. Original image courtesy digital cat@Flickr.

Abstract

Methods to undo the effects of motion blur are the subject of intense research, but evaluating and tuning these algorithms has traditionally required either user input or the availability of ground-truth images. We instead develop a metric for automatically predicting the perceptual quality of images produced by state-of-the-art deblurring algorithms. The metric is learned based on a massive user study, incorporates features that capture common deblurring artifacts, and does not require access to the original images (i.e., is “noreference”). We show that it better matches user-supplied rankings than previous approaches to measuring quality, and that in most cases it outperforms conventional full-reference image-similarity measures. We demonstrate applications of this metric to automatic selection of optimal algorithms and parameters, and to generation of fused images that combine multiple deblurring results.

Paper

PDF (32MB)

Source Code and Data

• A Matlab implementation. [zip]
• Data: images, user study raw data, fitted B-T score. [zip (840MB)]
• Data: example of images used in our user studies and case studies. [link to page]