A no-reference metric for evaluating the quality of motion deblurring

Abstract

Figure 1: 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. AbstractMethods 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.

Image Deconvolution Ringing Artifact Detection and Removal via PSF Frequency Analysis

“…The fact that ringing artifacts follow a sinusoidal pattern inspired the deconvolution artifact detection techniques of Zuo et. al [27] and Liu et al [28]. This is even more promising that from our PSF frequency analysis we can predict the frequency of this sinusoidal wave.…”

“…Three different types of blurry image and PSF pairs belong to this set. The first type of blurry images were created synthetically with known PSFs, as presented in [28]. An example of such images and PSFs is shown in Fig.…”

“…A quality metric that takes into account the deblurred image ringing artifacts through a non-reference scheme was recently introduced by Lui et al [28]. In order to provide an objective evaluation, this non-reference quality assessment was used in our experiments.…”

“…Since the employed filters are neither adapted to the content of the image nor to the PSF, high frequency patterns and textures in the image are falsely classified as artifacts in the filter response. A ringing detection procedure was proposed by Liu et al [28]. It consists of generating a pyramid of different scales of the recovered image and finding the gradient difference between each level of the pyramid.…”

No-reference image quality metric based on features evaluation

“…In [19], Liu et al designed a framework that used to measure the performance of deblurring algorithms. They made a collection of features, then select an optimal subset of these features based on a validation test.…”