Frequency Domain Blockiness and Blurriness Meter for Image Quality Assessment

Image and video compression introduces distortions (artefacts) to the coded image. The mostprominent artefacts added are blockiness and blurriness. Many existing quality meters arenormally distortion-specific. This paper proposes an objective quality meter for quantifying thecombined blockiness and blurriness distortions in frequency domain. The model first applies edgedetection and cancellation, then spatial masking to mimic the characteristics of the human visualsystem. Blockiness is then estimated by transforming image into frequency domain, followed byfinding the ratio of harmonics to other AC components. Blurriness is determined by comparing thehigh frequency coefficients of the reference and coded images due to the fact that blurrinessreduces the high frequency coefficients. Then, both blockiness and blurriness distortions arecombined for a single quality metric. The meter is tested on blocky and blurred images from theLIVE image database, with a correlation coefficient of 95-96%.