Correlation analysis for quality assessment of Region-of-Interest-coded images

摘要

Quality assessment is a technique used for evaluation of processes monitored by multimedia sources (Audio, Image, Video). Perceptual measures has been recently proposed for quality assessment of those sources. This work illustrates the correlation analysis of the Image Quality Assessment (IQA) required to evaluate a specific coding process. For this purpose, we create a database of images processed by Region-Of-Interest (ROI) quality preservation methods, namely ROI-Coders. Attained to this database, each processed image get a perceptual score value set of Mean-Observed-Scores (MOS) coming from a human-based test and those coming from Full-referenced perceptual measurements. As a result, two of the considered perceptual measures achieve the highest correlation values, suggesting those to be adequate for the IQA used in this experiment.