The Research For The Image Inpainting Quality Based On Human Visual Propety

With the image inpaiting algorithm improving, the assessment for the image quality repaired is extremely important. The human eyes are the ultimate recipient for the image information, the assessment for the repaired image quality tries to meet the human visual perception. This paper presents a full-reference evaluation and a no-reference evaluation method based on the human visual systerm.The full-reference evaluation establishes a human vision evaluation model, using Weber-Fechner law, the modulation transfer function process and the B-spline wavelets constructing the direction tunable filter to simulate the human visual systerm model. The experiment derives the result for the evaluation method has a strong correlation with subjective evaluation.For the image inpainted bluring, we proposed a no-reference image quality assessment base on the human visual property. Using the visual sensitivity function(CFS) for the human visual property to weight the noise images detected and get the value of the no-reference quality assessment consistenting with the subjective evaluation.