Perceptual error measure and its application to sampled and interpolated single-edged images

摘要: Error metrics quantify the difference between a reproduced image and corresponding unprocessed original image. A drawback of commonly used such as mean square error is their poor correlation with perceived quality We present framework for an alternative metric that uses distance in perceptual space to predict impairment images. The spanned by artifacts are introduced image-reproduction techniques. For reproduction using sampling interpolation it shown how multidimensional can be determined from sensory strengths artifacts’ periodic structure blur two orthogonal dimensions this space. In addition, we demonstrate that, after perceptual-error calibrated particular observer, successfully experimentally subjective sampled interpolated simple black-and-white