Abstract
A full-reference video quality assessment (VQA) method, called the ensemble-learning-based video quality assessment (EVQA) index, is proposed in this work. As compared with previous learning-based VQA methods, it has two unique features. First, EVQA adopts a frame-based learning mechanism to address the limited training data problem. Second, a dynamic image quality assessment(IQA) fusion scheme is developed by taking three factors into account: the spatial complexity and temporal context of a frame in a video source and the strength of IQA indices. In the test stage, EVQA applies the derived IQA fusion rule to different frames and take an average of the frame-based scores to generate the final video quality score. The superior performance of the proposed EVQA index is demonstrated by experimental results conducted on both LIVE and MCL-V video databases.