Reduced-reference picture quality estimation by using local harmonic amplitude information

In this paper, we present a reduced-reference picture quality assessment method for compressed digital images and video. Reduced reference refers to the availability of a preferably small subset of information extracted from an unprocessed, original picture. We propose local harmonic amplitude information computed from an edge-detected picture as the reduced-reference. By comparing the local harmonic amplitude information from both the original and the decoded (degraded) picture, we come up with a method employing gain and loss information which can be used to identify typical degradations such as blockiness or blur in the decoded picture. The proposed model can estimate picture quality even when the original, full reference picture is not present.