System and methods to measure noise and to generate picture quality prediction from source having no reference

摘要

Embodiments of the invention include a system and method to generate a quality rating prediction, such as MOS, of a video under test without using a second video as a reference. Instead of using a reference video, embodiments of the invention generate a pseudo reference video from measuring local impairments of the video under test. The local impairments are then accumulated to create a difference image. The difference image is then subtracted from the video under test to create the pseudo reference, to which the video under test is compared to generate the quality prediction rating. In some embodiments the local impairments may be measured by separating high frequency noise from the video under test from the video itself. Once separated, the noise is then measured and bandwidth compensated to generate the measured local impairments.

主权项

1. A method of generating a prediction of subjective picture quality from a single source of video input, comprising:
accepting one or more video frames at a single input; aggregating a series of local impairments to produce a local impairment difference image for each of the one or more video frames including:
measuring impairments of at least a portion of a current video frame to obtain a measured amount;adding the measured impairments to the current video frame to produce a video frame having added impairments; andsubtracting the video frame having the added impairments from the current video frame to produce the local impairment difference image; combining the local impairment difference image with a respective frame of the originally accepted one or more video frames to create a pseudo reference; and comparing the originally accepted one or more video frames with the pseudo reference to yield the prediction of subjective picture quality.