Abstract
This letter investigates a decision-fusion based image quality metric that performs multi-feature oriented method in contrast to individual feature as done by general methods. The comprehensive image quality feature is generated based upon the statistical method of Canonical Correlation Analysis (CCA), by which diverse quality features are incorporated. The efficient design of the proposed algorithm allows more accurate and stable quality prediction for complicated distortion, and another advantage over others is the flexibility of parameters tuning due to the employment of prior knowledge of the testing database.
