Perception-based video quality measurement system

• US 5,446,492 A
• Filed: 01/19/1993
• Issued: 08/29/1995
• Est. Priority Date: 01/19/1993
• Status: Expired due to Fees

• Current Assignee: United States Secretary of Commerce
• Original Assignee: United States Secretary of Commerce
• Inventors: Wolf, Stephen, Webster, Arthur A. III, Voran, Stephen D.
• Primary Examiner(s): Groody, James J.
• Assistant Examiner(s): GRANT, CHRISTOPHER C
• Application Number:
  US08/006,000
• Time in Patent Office: 952 Days
• Field of Search: 348/192, 348/193, 348/189, 348/180
• US Class Current: 348/192
• CPC Class Codes: H04N 17/00 Diagnosis, testing or measuring for television systems or their details
• CPC Class Codes: H04N 5/04 Synchronising for television systems using pulse code modulation H04N7/56
• CPC Class Codes: H04N 5/60 for the sound signals

Abstract

A method and apparatus for objectively measuring the image quality of a destination video signal generates measurement parameters that are indicative of human image quality perceptions. Subjective human test panel results are generated for a variety of test scenes and types of image impairment. Objective test results are also generated by the apparatus of the present invention for the variety of test scenes and image impairments. A statistical analysis means statistically analyzes the subjective and objective test results to determine operation of the apparatus. Accordingly, when the apparatus extracts test frames from the actual source and destination video signals and compares them, image quality parameters are output by the apparatus which are based on human image quality perceptions.

Claims

1. A quality measurement system for measuring an image quality of a destination video signal, comprising:

source feature extraction means for extracting source image features from a source video signal corresponding to the destination video signal;

destination feature extraction means for extracting destination image features from the destination video signal;

source time referencing means for providing a source time reference for each one of the source image features;

destination time referencing means for providing destination time reference for each one of the destination image features;

quality parameter generation means for generating image quality parameters, indicative of the image quality of the destination video signal based on the source video signal, from at least one source and at least one destination image feature.

2. The quality measurement system of claim 1 wherein the image quality parameters are indicative of human image quality perception.

5. The quality measurement system of claim 1, wherein the quality parameter generating means generates the image quality parameters from the source and destination image features and sub, active image quality test data.

6. The quality measurement system of claim 5, wherein the subjective image quality test data are indicative of human image quality perception.

7. The quality measurement system of claim 1, wherein quality parameter generation means generates the transmission video delay parameter based on at least the extracted source and destination image features.

8. The quality measurement system of claim 7, wherein the quality parameter generation means generates the transmission video delay parameter further based on the source and destination time references.

9. The quality measurement system of claim 1, wherein the image quality parameters include a transmission video delay parameter indicative of the time delay of the destination video signal.

10. The quality measurement system of claim 1, wherein the source feature extraction means comprises at least one source statistics processor and the destination extraction means comprises at least one destination statistics processor.

11. The quality measurement system of claim 1, wherein the source and destination image features each comprise first image features, the first image features being generated by source and destination spatial statistics processors from a source image frame and a destination image frame respectively.

17. The quality measurement system of claim 10, wherein the at least one source statistics processor and the at least one destination statistics processor each comprise one of a spatial statistics processor and a temporal statistics processor.

18. The quality measurement system of claim 1, wherein the source and destination image features each comprise second image features, the second image features being generated by source and destination temporal statistics processors from at least two source image frames and at least two destination image frames, respectively.

3. The quality measurement system of claim wherein the source feature extraction means comprises at least one source statistics processor and the destination feature extraction means comprises at least one destination statistics processor, wherein the at least one source and the at least one destination statistics processors operate based on statistical analysis of objective and subjective image quality test data.

4. The quality measurement system of claim 3, wherein the subjective image quality test data are indicative of human image quality perception.

12. A method for determining image quality of a destination video signal, comprising the steps of:

determining subjective image quality test data;

determining objective image quality test data;

generating source image features from a source video signal;

generating destination image features from the destination video signal;

generating source time reference data associated with each one of said source image features and destination time reference data associated with each one of said destination image features;

determining image quality parameters from the source and destination image features based on the subjective and objective image quality test data.

13. The method of claim 12, wherein the subjective image quality test data are indicative of human image quality perceptions.

14. A video quality measurement system for measuring a change in quality between a source video signal which includes a plurality of source data frames, each source data frame including a plurality of source pixels, and a destination video signal which includes a corresponding plurality of destination data frames, each destination data frame including a plurality of destination pixels, the video quality measurement system comprising:

first parameter extraction means for extracting a first plurality of parameters from the source video signal based on at least one objective measurement criteria;

second parameter extraction means for extracting a second plurality of parameters from the destination video signal based on the at least one objective measurement criteria;

alignment means coupled to the first and second parameter extraction means for performing time alignment between corresponding parameters of the first and second plurality of parameters;

combining means for combining the time aligned corresponding parameters of the first and second plurality of parameters to generate at least one quality indication;

the combining means combining the time aligned corresponding parameters of the first and second plurality of parameters according to a plurality of learned parameters, the plurality of learned parameters having been determined from a joint statistical analysis of a plurality of subjective viewer judgments.

15. The system of claim 14, wherein the first parameter extracting means and the second parameter extracting means each comprise:

Sobel filtering means for generating a plurality of Sobel filtered data frames, each Sobel filtered data frame corresponding to a respective data frame of the plurality of data frames and including a respective plurality of Sobel pixels; and

standard deviation means for determining a plurality of standard deviation values, each standard deviation value corresponding to a standard deviation of the Sobel pixels of a respective one of the plurality of Sobel filtered data frames.

16. The system of claim 14, wherein the first parameter extraction means and the second parameter extraction means each comprise:

absolute frame difference filtering means for generating a plurality of difference filtered data frames, each different filtered data frame corresponding to a respective data frame of the plurality of data frames and including a respective plurality of difference filtered pixels; and

temporal statistics generating means for determining a plurality of temporal statistic values, each temporal statistic value corresponding to the temporal difference of the difference filtered pixels of a respective one of the plurality of difference filtered data frames.