METHOD FOR DERIVING NOISE STATISTICAL PROPERTIES OF A SIGNAL

• US 20070237215A1
• Filed: 02/28/2007
• Published: 10/11/2007
• Est. Priority Date: 03/29/2006
• Status: Active Grant

• Current Assignee: Sony Deutschland GmbH (Sony Group Corp.)
• Original Assignee: Sony Deutschland GmbH (Sony Group Corp.)
• Inventors: SARTOR, Piergiorgio
• Granted Patent: US 7,876,975 B2
• Time in Patent Office: Days
• US Class Current: 375/224
• CPC Class Codes: G06T 5/003 Deblurring; Sharpening vibration or motion blur correction for cameras comprising an electronic image sensor H04N23/682
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening
• CPC Class Codes: G10L 21/0208 Noise filtering

Abstract

The invention relates to a method for deriving noise statistical properties (NSP) for a signal (S) wherein first distorted noise statistical properties (DNSP) are derived from a distorted signal (DS). In addition the distorted noise statistical properties (DNSP) are subjected to a deconvolution/inversion process (S3) to thereby derive un-distorted noise statistical properties (NSP) for said underlying un-distorted signal (S).

Claims

1. Method for deriving noise statistical properties (NSP) for a signal (S),

a) wherein a distorted signal (DS) originating from an underlying signal (S) to be evaluated is provided (S1),

(b) wherein distorted noise statistical properties (DNSP) are derived (S2) from said distorted signal (DS),

(c) wherein said distorted noise statistical properties (DNSP) are subjected to a inversion process (S3) with respect to a respective noise model (NM) for said underlying signal (S) to be evaluated, and

(d) wherein undistorted noise statistical properties are derived (S4) as said noise statistical properties (NSP) for said underlying undistorted signal (S).

2. Method according to claim 1,

wherein a deconvolution and/or said inversion process (S3) is designed for removing the effect of distortion in particular with respect to said distorted noise statistical properties (DNSP).

3. Method according to any one of the preceding claims,

wherein said distortion is given as at least one of the group consisting of a distortion function and a distortion distribution.

4. Method according to any one of the preceding claims,

wherein said distortion is given as at least one of the group consisting of clipping and filtering.

5. Method according to any one of the preceding claims 3 or 4,

wherein said distortion function has a unique invers distortion function.

6. Method according to any one of the preceding claims,

wherein at least one of the group consisting of said distorted noise statistical properties (DNSP), said undistorted noise statistical properties, said noise statistical properties (NSP) and a part thereof is given as respective distorted noise statistical data (DNSD), undistorted noise statistical data and noise statistical data (NSD), respectively.

7. Method according to claim 6,

wherein at least one of said undistorted noise statistical data (DNSD), undistorted noise statistical data, and noise statistical data (NSD) comprise at least one of the group consisting of statistical mean and statistical variance.

8. Method according to any one of the preceding claims,

wherein said noise model (NM) is given by at least one of the group consisting of noise model function and noise model distribution.

9. Method according to claim 8,

wherein said noise model function have a corresponding unique inverse noise model function.

10. Method according to any one of the preceding claims,

wherein said noise model function is a Gaussian type function.

11. Method according to any one of the preceding claims,

wherein said process of deconvolution/inversion (S3) is at least in part performed analytically.

12. Method according to any one of the preceding claims,

wherein said process of deconvolution/inversion (S3) is at least in part performed numerically.

13. Method according to claim 12,

wherein said process of deconvolution/inversion (S3) is at least in part performed iteratively.

14. Method according to claim 13,

(e) wherein at least one of the group consisting of a mean value and a variance value is first derived or measured from said and for said distorted signal (DS) as a distorted mean value (DMV) and as a distorted variance value (DVV), respectively,

(f) wherein then said measured mean value (DMV) and said measured variance value (DVV) are initially set and used as starting iteration values and as intermediate iteration values,

(g) wherein then for the next iteration step based on one of said intermediate iteration values for said variance and said mean the other one of said values for said variance and said mean is evaluated by numerical inversion and taken as a new respective intermediate iteration value for the other one of said variance and said mean, respectively,

(h) wherein then based on the other one of said intermediate iteration values for said variance and said mean the one of said values for said variance and said mean is evaluated by numerical inversion and taken as a new respective intermediate iteration value for the one of said variance and said mean, respectively,

(j) wherein at least one of said steps (g) and (h) is repeated for the next iteration step until given stopping criteria are fulfilled, and

(k) wherein after stopping the iteration the recently evaluated and calculated iteration values for said variance and for said mean are taken as said noise statistical data (NSD) or as a part or pre-form thereof.

15. Method according to claim 14,

wherein said stopping criteria are referred to as being fulfilled if the differences of the values for the variance and for the mean of directly succeeding iterations or iteration steps are below given respective threshold values.

16. Method according to any one of the preceding claims,

wherein a look-up table (LUT) is formed representing a one-to-one relationship between said distorted noise statistical data (DNSP, DNSD) and said respective non-distorted noise statistical data (NSP, NSD).

17. System, device, or apparatus for deriving noise statistical properties (NSP) for a signal (S),

which is adapted and which comprises means for carrying out a method for deriving noise statistical properties (NSP) for a signal (S) according to any one of the claims 1 to 16 and the steps thereof.

18. Computer program product,

comprising computer program means which is adapted and arranged in order to perform a method for deriving noise statistical properties (NSP) for a signal (S) according to any one of the claims 1 to 16 and the steps thereof, when it is executed on a computer or a digital signal processing means.

19. Computer readable storage medium,

comprising a computer program product according to claim 18.