No-reference image-quality metric based on blur radius and visual blockiness

摘要

Abstract. We propose new blur and blocking metrics and then presentna no-reference image-quality assessment method using these blur andnblocking metrics. To compute the blur metric, we first estimated a blurnradius from a given image and its reblurred version by using edge differ-nences and edge amplitudes. Because blurring in edge regions is gener-nally more sensitive to human perception, the blur metric was estimatednfrom the edge blocks. We also used kurtosis and structural similarity tonbetter estimate the blur metric. To compute the blocking metric, thenblocking artifact was modeled as a 2-D step function and the blockinessnvisibility was estimated by the brightness difference between adjacentnblocks. After the blocky position was determined, the blocking metric wasncomputed from the six differences between four adjacent blocks. Experi-nmental results show that the objective quality scores correlated highlynwith the subjective quality scores. © 2010 Society of Photo-Optical InstrumentationnEngineers. u0001DOI: 10.1117/1.3366671