Gabor difference analysis of digital video quality

The rapid increase in the development of digital video systems has generated a strong need for objective video quality metrics. Two methods are presented in this work. One is the Gabor difference analysis (GDA) full reference method, and the other is the reverse frame prediction (RFP) no reference video quality method. Both methods are based on the multi-channel properties of the human visual system (HVS). Gabor filtering is used in both methods. In the GDA method, a reference and a degraded digital video sequence are compared by taking into account various psycho-perceptual properties of the HVS. The RFP method does not require a reference video stream and is intended for in-service testing and on-line monitoring. The performances of the proposed methods are evaluated. The methods in this work are shown to be consistent with the data from subjective testing over a wide range of scenes. This work is critical to the evaluation of the effectiveness of compression schemes on HDTV imagery.