ut ut

Laboratory for Image & Video Engineering


LIVE Image Quality Assessment Database Algorithm Comparison Results


The LIVE IQA database, has been available for IQA R&D engineers and scientists since 2003, has served as a response for developing many of the worlds's leading image quality models.


Image Quality Assessment research strongly depends upon the results of subjective experiments to provide calibration data as well as a testing mechanism. After all, the goal of all IQA research is to make quality predictions that are in agreement with the subjective opinion of human observers.

As an additional service, LIVE is making available lists of reported results in three categories: Reference (including Reduced-Reference), No-Refecence and Completely Blind. Each list can be sorted by one of four metrics: Root Mean-Squared Error (RMSE), Linear (Pearson) Correlation (LCC), Spearman's Rank-Order Correlation (SROCC), and Outlier Ratio (OR). For each metric, ordering of an algorithm will only occur if that metric was reported by the authors. Ordering by metric can be achieved by clicking on the appropriate column header.

Since scientific reproducibility is of paramount performance, each category has two lists: a primary list, where executable or easily-compiled source code is available - so the community may verify and duplicate results, and a secondary list, of algorithms with reported results but no supplied code. If code is made available by a public link, then it may be moved from the secondary list to the primary list. No reason for not supplying code (proprietary, etc) will be entertained as a substitute for entry to the primary list.

Category lists:

Back to Subjective Image Quality Database page