Stereo matching for 3D encoding and quality assessment

摘要

Systems and methods may be provided embodying a novel approach to measuring degradation (or distortion) by analyzing disparity maps from original 3D video and reconstructed 3D video. The disparity maps may be derived using a stereo-matching algorithm exploiting 2-view stereo image disparity. An overall distortion measure may also be determined resulting from the weighted sum of plural measures of distortions, one of the plural distortion measures corresponding to a measure of disparity degradation, and another one corresponding to a measure of geometrical distortion. The measure (or overall distortion measure) is used during real-time encoding to effect various decisions, including mode decision in the coding of each corresponding stereo pair, and in rate control (including stereo pair quantization).

主权项

1. A method comprising:
receiving a source 3D video stream; receiving or generating disparity information associated with the source 3D video stream; receiving a distorted version of the same source 3D video stream; generating disparity information associated with the distorted 3D video stream by employing a stereo matching algorithm to derive a disparity between a left view and a right view of the source 3D video stream; comparing the disparity information associated with the source 3D video stream with the disparity information associated with the distorted 3D video stream; and altering an encoding process based on the results of comparing the disparity information associated with the source 3D video stream with the disparity information associated with the reconstructed and distorted 3D video stream.