LAUSR

When individuals view an image, they form an opinion by considering only some parts of the image, which are defined as regions of interest. Based on this hypothesis, in this paper, a local gray-scale fluctuation measurement is used to select image patches with rich structural information and texture that can attract human visual perception. First, a gray-scale fluctuation map (GFM) is calculated based on a specific primitive. The pixel values in the GFM reflect the gray-scale fluctuation level of the pixel in the same location in the corresponding image. Second, the patches are selected via the gray-scale fluctuation conditions in the GFM. Third, existing image quality assessment (IQA) methods are used to measure the quality of each patch. Additionally, the quality assessment results of each patch are combined as the quality score of the entire image. The experimental results on five open databases demonstrate that the proposed method has a positive effect on IQA for existing methods. Additionally, the proposed method can improve the prediction accuracy of the quality assessment of diversely distorted types.