7 151 801

7,151,801 Title:

Method and system for enhancing data quality

The quality (e.g., resolution) of image data, video data, and sound data representing a scene or signal is enhanced by a quality enhancement function trained on high quality and low quality representations of a portion of the same scene or signal. The enhancement function is thus optimized to the most relevant input. A training algorithm uses low quality image of a scene or scene portion, along with a high quality image of the same scene or scene portion, to optimize the parameters of a quality enhancement function. The optimized enhancement function is then used to enhance other low quality images of the scene or scene portion. Sound data is enhanced by using a low resolution sample of a portion of a signal, and a high resolution sample of the same signal portion, to train a quality enhancement function which is then used to enhance the remainder of the signal.

We claim:

1. A method for enhancing data quality using a first data set, a second data set, and a quality enhancement function having at least one parameter, the first data set representing atleast one of a scene and a signal, the second data set representing a first sub-portion of the at least one of the scene and the signal, the first data set having a first amount of quality, the second data set having a second amount of quality, thesecond amount of quality being greater than the first amount of quality, the first data set including first and second data subsets, the first data subset representing the first sub-portion of the at least one of the scene and the signal, the second datasubset representing a second sub-portion of the at least one of the scene and the signal, each of the first and second data subsets having the first amount of quality, the method comprising: processing the first data subset and the second data set by atraining algorithm to determine a respective learned value of each one of the at least one parameter, wherein the quality enhancement function is operable to derive, based on the first data subset and using the respective learned value of each one of theat least one parameter, a data set approximating the second data set; and processing the second data subset by the quality enhancement function using the respective learned value of each one of the at least one parameter to derive a third data set, thethird data set representing the second sub-portion of the at least one of the scene and the signal, the third data set having a third amount of quality, the third amount of quality being greater than the first amount of quality.

2. A method according to claim 1, wherein the first amount of quality comprises a first resolution, the second amount of quality comprising a second resolution, the third amount of quality comprising a third resolution.

3. A method according to claim 2, wherein the at least one of the scene and the signal comprises the scene, the first resolution comprising at least one of a first spatial resolution, a first intensity resolution, a first spectral resolution,and a first polarization resolution, the second resolution comprising at least one of a second spatial resolution, a second intensity resolution, a second spectral resolution, and a second polarization resolution, the third resolution comprising at leastone of a third spatial resolution, a third intensity resolution, a third spectral resolution, and a third polarization resolution.

4. A method according to claim 2, wherein the first resolution comprises a first temporal resolution, the second resolution comprising a second temporal resolution, the third resolution comprising a third temporal resolution.

5. A method according to claim 2, wherein the at least one of the scene and the signal comprises the signal, the first resolution comprising at least one of a first signal level resolution and a first temporal resolution, the second resolutioncomprising at least one of a second signal level resolution and a second temporal resolution, the third resolution comprising at least one of a third signal level resolution and a third temporal resolution.

6. A method according to claim 1, wherein the first amount of quality comprises a first signal-to-noise ratio, the second amount of quality comprising a second signal-to-noise ratio, the third amount of quality comprising a thirdsignal-to-noise ratio.

7. A method according to claim 1, wherein the first amount of quality comprises a first accuracy, the second amount of quality comprising a second accuracy, the third amount of quality comprising a third accuracy.

8. A method according the claim 1, wherein the quality enhancement function comprises a polynomial function, the at least one parameter comprising one or more coefficients of the polynomial function.

9. A method according to claim 8, wherein the training algorithm comprises determining a respective value of each of the one or more coefficients which minimizes a sum of square differences between the second data set and the data setapproximating the second data set.

10. A method according to claim 1, wherein the training algorithm comprises determining a respective value of each one of the at least one parameter which minimizes a sum of square differences between the second data set and the data setapproximating the second data set.

11. A method for enhancing data quality using a high quality image data set representing a scene, a first low quality image data set representing the scene, a second low quality image data set representing the scene, and a quality enhancementfunction having at least one parameter, an image corresponding to the high quality image data set having a first amount of quality, images corresponding to the first and second low quality image data sets having a second amount of quality, the firstamount of quality being greater than the second amount of quality, the method comprising: processing the high quality image data set and the first low quality image data set by a training algorithm to determine a respective learned value of each one ofthe at least one parameter, wherein the quality enhancement function is operable to derive, based on the first low quality image data set and using the respective learned value of each one of the at least one parameter, an image data set approximatingthe high quality image data set; and processing the second low quality image data set by the quality enhancement function using the respective learned value of each one of the at least one parameter to derive a data set corresponding to an image havinga third amount of quality, the third amount of quality being greater than the second amount of quality.

12. A method according to claim 11, wherein the first amount of quality comprises a first resolution, the second amount of quality comprising a second resolution, the third amount of quality comprising a third resolution.

13. A method according to claim 12, wherein the first resolution comprises at least one of a first spatial resolution, a first intensity resolution, a first spectral resolution, and a first polarization resolution, the second resolutioncomprising at least one of a second spatial resolution, a second intensity resolution, a second spectral resolution, and a second polarization resolution, the third resolution comprising at least one of a third spatial resolution, a third intensityresolution, a third spectral resolution, and a third polarization resolution.

14. A method according to claim 11, wherein the first amount of quality comprises a first signal-to-noise ratio, the second amount of quality comprising a second signal-to-noise ratio, the third amount of quality comprising a thirdsignal-to-noise ratio.

15. A method according to claim 11, wherein the first amount of quality comprises a first accuracy, the second amount of quality comprising a second accuracy, the third amount of quality comprising a third accuracy.

16. A method according the claim 11, wherein the quality enhancement function comprises a polynomial function, the at least one parameter comprising one or more coefficients of the polynomial function.

17. A method according to claim 16, wherein the training algorithm comprises determining a respective value of each of the one or more coefficients which minimizes a sum of square differences between the high quality image data set and theimage data set approximating the high quality image data set.

18. A method according to claim 11, wherein the training algorithm comprises determining a respective value of each one of the at least one parameter which minimizes a sum of square differences between the high quality image data set and theimage data set approximating the high quality image data set.

19. A method according the claim 11, wherein the high quality image data set represents the scene as viewed from a first viewpoint, the first low quality image data set representing the scene as viewed from a second viewpoint, the second lowquality image data set representing the scene as viewed from a third viewpoint, the second and third viewpoints being proximate to the first viewpoint.

20. A system for enhancing data quality using a first data set, a second data set, and a quality enhancement function having at least one parameter, the first data set representing at least one of a scene and a signal, the second data setrepresenting a first sub-portion of the at least one of the scene and the signal, the first data set having a first amount of quality, the second data set having a second amount of quality, the second amount of quality being greater than the first amountof quality, the first data set including first and second data subsets, the first data subset representing the first sub-portion of the at least one of the scene and the signal, the second data subset representing a second sub-portion of the at least oneof the scene and the signal, each of the first and second data subsets having the first amount of quality, the system comprising a processing arrangement configured to perform the steps of: processing the first data subset and the second data set by atraining algorithm to determine a respective learned value of each one of the at least one parameter, wherein the quality enhancement function is operable to derive, based on the first data subset and using the respective learned value of each one of theat least one parameter, a data set approximating the second data set; and processing the second data subset by the quality enhancement function using the respective learned value of each one of the at least one parameter to derive a third data set, thethird data set representing the second sub-portion of the at least one of the scene and the signal, the third data set having a third amount of quality, the third amount of quality being greater than the first amount of quality.

21. A system according to claim 20, wherein the first amount of quality comprises a first resolution, the second amount of quality comprising a second resolution, the third amount of quality comprising a third resolution.

22. A system according to claim 21, wherein the at least one of the scene and the signal comprises the scene, the first resolution comprising at least one of a first spatial resolution, a first intensity resolution, a first spectral resolution,and a first polarization resolution, the second resolution comprising at least one of a second spatial resolution, a second intensity resolution, a second spectral resolution, and a second polarization resolution, the third resolution comprising at leastone of a third spatial resolution, a third intensity resolution, a third spectral resolution, and a third polarization resolution.

23. A system according to claim 21, wherein the first resolution comprises a first temporal resolution, the second resolution comprising a second temporal resolution, the third resolution comprising a third temporal resolution.

24. A system according to claim 21, wherein the at least one of the scene and the signal comprises the signal, the first resolution comprising at least one of a first signal level resolution and a first temporal resolution, the secondresolution comprising at least one of a second signal level resolution and a second temporal resolution, the third resolution comprising at least one of a third signal level resolution and a third temporal resolution.

25. A system according to claim 20, wherein the first amount of quality comprises a first signal-to-noise ratio, the second amount of quality comprising a second signal-to-noise ratio, the third amount of quality comprising a thirdsignal-to-noise ratio.

26. A system according to claim 20, wherein the first amount of quality comprises a first accuracy, the second amount of quality comprising a second accuracy, the third amount of quality comprising a third accuracy.

27. A system according the claim 20, wherein the quality enhancement function comprises a polynomial function, the at least one parameter comprising one or more coefficients of the polynomial function.

28. A system according to claim 27, wherein the training algorithm comprises determining a respective value of each of the one or more coefficients which minimizes a sum of square differences between the second data set and the data setapproximating the second data set.

29. A system according to claim 20, wherein the training algorithm comprises determining a respective value of each one of the at least one parameter which minimizes a sum of square differences between the second data set and the data setapproximating the second data set.

30. A system for enhancing data quality using a high quality image data set representing a scene, a first low quality image data set representing the scene, a second low quality image data set representing the scene, and a quality enhancementfunction having at least one parameter, an image corresponding to the high quality image data set having a first amount of quality, images corresponding to the first and second low quality image data sets having a second amount of quality, the firstamount of quality being greater than the second amount of quality, the system comprising a processing arrangement configured to perform the steps of: processing the high quality image data set and the first low quality image data set by a trainingalgorithm to determine a respective learned value of each one of the at least one parameter, wherein the quality enhancement function is operable to derive, based on the first low quality image data set and using the respective learned value of each oneof the at least one parameter, an image data set approximating the high quality image data set; and processing the second low quality image data set by the quality enhancement function using the respective learned value of each one of the at least oneparameter to derive a data set corresponding to an image having a third amount of quality, the third amount of quality being greater than the second amount of quality.

31. A system according to claim 30, wherein the first amount of quality comprises a first resolution, the second amount of quality comprising a second resolution, the third amount of quality comprising a third resolution.

32. A system according to claim 31, wherein the first resolution comprises at least one of a first spatial resolution, a first intensity resolution, a first spectral resolution, and a first polarization resolution, the second resolutioncomprising at least one of a second spatial resolution, a second intensity resolution, a second spectral resolution, and a second polarization resolution, the third resolution comprising at least one of a third spatial resolution, a third intensityresolution, a third spectral resolution, and a third polarization resolution.

33. A system according to claim 30, wherein the first amount of quality comprises a first signal-to-noise ratio, the second amount of quality comprising a second signal-to-noise ratio, the third amount of quality comprising a thirdsignal-to-noise ratio.

34. A system according to claim 30, wherein the first amount of quality comprises a first accuracy, the second amount of quality comprising a second accuracy, the third amount of quality comprising a third accuracy.

35. A system according the claim 30, wherein the quality enhancement function comprises a polynomial function, the at least one parameter comprising one or more coefficients of the polynomial function.

36. A system according to claim 35, wherein the training algorithm comprises determining a respective value of each of the one or more coefficients which minimizes a sum of square differences between the high quality image data set and theimage data set approximating the high quality image data set.

37. A system according to claim 30, wherein the training algorithm comprises determining a respective value of each one of the at least one parameter which minimizes a sum of square differences between the high quality image data set and theimage data set approximating the high quality image data set.

38. A system according the claim 30, wherein the high quality image data set represents the scene as viewed from a first viewpoint, the first low quality image data set representing the scene as viewed from a second viewpoint, the second lowquality image data set representing the scene as viewed from a third viewpoint, the second and third viewpoints being proximate to the first viewpoint.