No-reference quality assessment of blurred frames

Visual Quality Assessment (VQA) methods applicable to evaluate noise, blurring, and MPEG distortions may be considered as an extension of Image Quality Assessment (IQA) methods. At the same time, the VQA is a special issue for studying in a sense of available set of closing frames. Given our interest in evaluation of frame blurring, we focus the discussion on the no-reference methods regarding to the main types of blurriness occurring in video sequences, such as the motion blur, defocus blur, shaking blur, and atmospheric blur. First, we detect the blurred regions in frame using a simple procedure of imposed blurring. Second, a quality degree of blurriness is evaluated according to the defined type. The obtained estimators facilitate the initial level for deblurring algorithms in order to decide should be the blurred frames restored or removed from the following processing. The dataset Sports Videos in the Wild (SVW) provides the rich test material for experiments. The resultant estimators coincide well with perceptual Human Visual System (HVS) evaluation. (C) 2018 The Authors. Published by Elsevier Ltd.