On Blind Quality Assessment of JPEG Images
Guangtao Zhai · 2016
JPEG is still the most widely used image compression format. Perceptual quality assessment for JPEG images has been extensively studied for the last two centuries. While a large number of no-reference perceptual quality metrics have been proposed along the years, it is shown in this paper that on existing image quality databases, statistically, performance of many those metrics is not better than the quality factor (Q) for JPEG images, as used in the popular implementation by the IJG (Independent JPEG Group). It should be noted that Q or the quantization table computed from Q is almost always availabe at the decoder end so analysis is focused on no-reference or blind quality assessment metrics. This research highlights the fact that despite of many progress achieved in that area, JPEG quality assessment is still a topic worth revisiting and further investigation.