No-reference image blurriness assessment using divisive normalization

During image processing, it is observed that the input images got distorted for various reasons. The distortion of images lowers the quality of the images, which affects the processing of the images. Therefore, assessment of the quality of an image is very much necessary before further processing it. Blurriness is the most frequent form of degradation in images. Hence, the quality analysis of an image by quantifying its blurriness is an attractive area of research. Although, image quality assessment has three major types, namely full reference, reduced reference, and no reference, the third type of image quality assessment technique has the most applicability. A lot of research has been conducted on this topic. We surveyed and classified those according to their computation strategy. We found a new observation for the blurred images when they are transformed through divisive normalized transformation. We have used this observation to propose a novel image quality metric named no-reference image blurriness estimation metric (NIBEM). We have provided the mathematical explanation behind the observation. We evaluated the metric with public image quality databases and compared the performance with the state-of-the-art metrics. The comparison proves the effectiveness of our proposed metric. Thus, this paper includes three contributions as a whole-survey on no-reference blur metric, mathematical and experimental analysis of the behavioral changes in the distribution of DNT coefficients based on image blurriness, and an image quality metric NIBEM. The authors of the paper believe that the contributions enrich the research in image quality assessment.

Automated summary

This paper presents a study on the no-reference blur metric in image quality assessment, proposing a new metric called NIBEM. The effectiveness of the metric is demonstrated through evaluation with public image quality databases and comparison with other metrics.