A Pixel-Based Framework for Blind Image Quality Assessment

Abstract

Image quality assessment delivers an objective (algorithmic) score. This score must be correlated as much as possible to the human visual perception when seeing images distorted by some compression and/or transmission tools, e.g. JPEG compression, white noise, fast fading. The process can be a full-reference or a null-reference image quality assessment depending on if it has the knowledge of the original undistorted image. This paper presents a prospective for building a null-reference image quality assessment, for a given distorted image, its aim is to deliver both the global objective score and the distortion map where each pixel has a distortion measurement.