Human perception related distortion metrics for blockiness and blur compression artefacts

Objective image and video quality evaluations generally provide an overall measure of distortions. Parameter optimization and benchmarking of codecs require evaluating individual distortions. There are not many objective metrics that distinguish individual compression artefacts which are rapid to evaluate and correlate with human perception. The author of the paper presents the correlation that had been established between the metrics of compression artefact distortions that objectively and separately measure blockiness and blur artefacts introduced by image compression codecs. The methodology and the approach use synthetic test patterns specially designed for each of the two artefacts. Evaluation of quality in multimedia applications requires that the distortion metrics be correlated to human perception. One limitation of the previously proposed two metrics was that no correlation was established to human perception. This paper investigates this correlation, using SSIM as a proxy for human perception, demonstrating a strong correlation.