Video quality analysis using a linear approximation technique
View Patent ↗A video analysis technique includes correlating frames from a processed video with frames from a pre-processed, original video. A linear approximation of a relationship between the correlated frames is determined. A disclosed example includes determining a linear approximation that maximizes the number of processed video frames that fit into the linear approximation. The linear approximation and whether any frames do not fit within the linear approximation is then used to provide quality information for analyzing a quality of the processed video.
1. A method of analyzing a processed video, comprising
correlating each frame of the processed video to a corresponding frame of a pre-processed version of the video;
determining a linear approximation from a line with respect to a relationship between the correlated frames; and
determining a quality of the processed video based upon whether any of the frames does not fit within the linear approximation.
2. The method of claim 1 , comprising
assigning a watermark to each frame of the pre-processed version of the video before it is processed;
determining the watermark of each of the processed video frames; and
using video frames of the processed video and the pre-processed video having the same watermark as the correlated frames.
3. The method of claim 1 , comprising
determining image edge information for each of the pre-processed video frames;
determining processed image edge information of each of the processed video frames; and
using video frames having corresponding image edge information as the correlated frames.
4. The method of claim 3 , comprising
determining a correspondence between
(i) the processed image edge information of each one of the processed video frames and
(ii) the image edge information of all of the pre-processed video frames; and
correlating each processed video frame with the one of the pre-processed video frames having the greatest correspondence of image edge information.
5. The method of claim 1 , comprising
arranging the processed video frames in sequential order along one axis;
arranging the correlated pre-processed video frames in sequential order along another axis perpendicular to the one axis; and
determining the linear approximation from a line relative to the axes that includes as many pairs of the correlated frames as possible.
6. The method of claim 5 , comprising
determining whether any frame is on the line as an indication of whether any processed video frame includes an error.
7. The method of claim 1 , comprising
selecting a plurality of groups of correlated frame pairs;
determining a respective linear relationship for each of the plurality of groups such that an index of the pre-processed frames in the group is equal to a frame offset plus an index of the processed frames in the group multiplied by a compensation rate.
8. The method of claim 7 , comprising
determining the linear approximation from a line having the one of the offsets and compensation rates from the respective determined linear relationships that occurs most often among the respective linear relationships.
9. The method of claim 8 , comprising
determining whether the frame index of any processed video frame is on the line as an indication of whether any processed video frame includes an error.
10. The method of claim 1 , comprising
determining a compensation rate of the processed video based upon a slope of the linear approximation.
11. The method of claim 10 , wherein the linear approximation includes a line having a slope of unity.
12. The method of claim 1 , comprising
determining a frame index offset between the pre-processed and processed video frames; and
determining a lag of the processed video relative to the pre-processed video from the determined offset.
13. The method of claim 1 , comprising
establishing the linear approximation to include as many of the frames as possible on a single line that describes the relationship between the correlated frames; and
determining whether any frame is on the line as in indication of whether any processed video frame includes an error.
14. The method of claim 1 , comprising
determining a number of frames that do not fit within the linear approximation as an indicator of quality.
15. The method of claim 14 , comprising
determining a percentage of the frames that do not fit within the linear approximation as an indicator of quality.
16. The method of claim 1 , comprising
determining a distribution of any frames that do not fit within the linear approximation as an indicator of quality.
17. A device for analyzing a processed video, comprising
a correlation portion configured to correlate each frame of the processed video to a corresponding frame of a pre-processed version of the video;
a linear approximation portion configured to determine a linear approximation from a line with respect to a relationship between the correlated frames; and
a quality determining portion configured to determine a quality of the processed video based upon whether any of the frames does not fit within the linear approximation.
18. The device of claim 17 , wherein the correlation portion is configured to
correlate the frames based upon whether a watermark associated with a processed video frame corresponds to a watermark of any of the pre-processed video frames; or
correlate the frames based upon how closely image edge information for each of the processed video frames corresponds to image edge information for the pre-processed video frames.
19. The device of claim 17 , wherein the linear approximation portion is configured to determine a linear approximation that maximizes a number of the frames that fits within the linear approximation.
20. The device of claim 17 , wherein the quality determining portion is configured to determine at least one of
a number of frames that do not fit within the linear approximation,
a percentage of frames that do not fit within the linear approximation, or
a distribution of frames that do not fit within the linear approximation
as an indicator of quality.