A no-reference image blurriness metric in the spatial domain

This paper presents a no-reference image quality metric for Gaussian blurred image. The metric is implemented in the spatial domain without the need of data conversion or training. It utilizes the just noticeable blur (JNB) model to estimate the amount of blurriness at each edge in the image. Then the probability distribution histogram of edge blurriness is built to calculate the final blur metric of entire image. The performance of the metric is demonstrated by comparing it with existing no-reference blurriness metrics. And the experimental results on multiple image quality assessment databases show that the proposed metric is highly consistent with the subjective quality evaluations.