Apparatus and methods for image and signal processing

• US 20020150302A1
• Filed: 03/18/2002
• Published: 10/17/2002
• Est. Priority Date: 07/31/1997
• Status: Active Grant

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: McCarthy, Sean T., Owen, William G.
• Granted Patent: US 7,046,857 B2
• Time in Patent Office: Days
• US Class Current: 382/254
• CPC Class Codes: G06T 2207/20048 Transform domain processing
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06T 9/00 Image coding bandwidth or redundancy reduction for static pictures H04N1/41; coding or decoding of static colour picture signals H04N1/64; methods or arrangements for coding, decoding, compressing or decompressing digital video signals H04N19/00

Abstract

An apparatus and methods for efficiently processing signal and image data are described. The invention provides a representation of signal and image data that can be used as a figure of merit to compare and characterize different signal processing techniques. The representation can be used as an intermediate result that is may be subjected to further processing, and/or may be used as a control element for processing operations. As a provider of an intermediate result, the invention can be used as a step in processes for the transduction, storage, enhancement, refinement, feature extraction, compression, coding, transmission, or display of image, audio and other data. The invention improves manipulation of data from intrinsically unpredictable, or partially random sources. The result is a concise coding of the data in a form permitting robust and efficient data processing, a reduction in storage demands, and restoration of original data with minimal error and degradation. The invention provides a system of coding source data derived from the external environment, whether noise-free or contaminated by random components, and regardless of whether the data are represented in its natural state, such as photons, or have been pre-processed.

Claims

1. A method of processing a set of input data X(v) representing a desired signal component plus an undesired contamination component, the method comprising the following steps:

obtaining an ensemble-averaged power spectrum of the signal component, <

|K_S(v)|²>

;

obtaining an ensemble-averaged power spectrum of the contamination component, <

|K_N(v)|²>

;

forming a term B²=<

|K_N(v)|²>

/<

|K_S(v)|²>

;

forming a filter function W(v), where |W(v)|=[1+b²B(v)²]<sup>−

1</sup>, and b is greater than zero;

forming a term U(v), where |U(v)|=[W(v)(1−

W(v))]^½

; and

processing X(v) to form a result U(v) X(v).

2. The method of claim 1, wherein the step of obtaining an ensemble-averaged power spectrum of the signal component, further comprises:

averaging a set of data known to represent the signal component of the set of input data.

3. The method of claim 1, wherein the step of obtaining an ensemble-averaged power spectrum of the contamination component, further comprises at least one step selected from a group consisting of (i) assuming a model for the contamination component and determining its ensemble-averaged power spectrum, and (ii) assuming a model for the contamination component that is a representation of white noise.

4. The method of claim 1, further comprising at least one step selected from a group consisting of (i) further processing the result U(v) X(v) by applying a desired signal processing technique, and (ii) quantizing the result U(v) X(v).

5. The method of claim 1, further including the step of quantizing the result U(v) X(v), wherein said quantizing includes at least one step selected from a group consisting of (i) comparing the result to a set of reference values determined from the result'"'"'s ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values, and (ii) comparing the result to a set of reference values determined from a model of an ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values.

6. The method of claim 1, wherein b² is inversely proportional to [a₁+Iⁿ], where a₁ is a constant, I is a mean value of the signal, and n is an integer.

7. The method of claim 1, wherein the set of input data is representative of visual image data.

8. A method of processing data representative of visual images, the method comprising the following steps:

forming a filter function W(v), where |W(v)|=[1+b²B(v)²]<sup>−

1</sup>, b² is a constant selected to satisfy |W(v)|<

1 for all v, and B(v) is proportional to v;

forming a term U(v), where |U(v)|=[W(v)(1−

W(v))]^½

; and

processing X(v) to form result U(v) X(v).

9. The method of claim 8, further comprising at least one step selected from a group consisting of (i) further processing the result U(v) X(v) by applying a desired signal processing technique, and (ii) providing the result U(v) X(v) as a control term to vary amplitude spectrum of the function filter function W(v).

10. The method of claim 9, further comprising at least one step selected from a group consisting of (i) quantizing the result U(v) X(v), (ii) quantizing the result by comparing the result to a set of reference values determined from the result'"'"'s ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values, and (iii) quantizing the result U(v) X(v) by comparing the result to a set of reference values determined from a model of an ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values.

11. The method of claim 9, wherein b² is inversely proportional to [a₁+Iⁿ], where a₁ is a constant, I is a mean value of the signal, and n is an integer.

13. The method of claim 12, wherein input data X is representative of visual image data.

14. The method of claim 12, including a step of further processing the result U(v) X(v) using at least one technique selected from a group consisting of (i) applying a desired signal processing technique, (ii) providing the result U(v) X(v) as a control term to vary amplitude spectrum of the function filter function W(v), (iii) quantizing the result U(v) X(v), (iv) quantizing the result U(v) X(v) by comparing the result to a set of reference values determined from the result'"'"'s ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values, and (v) quantizing the result U(v) X(v) by comparing the result to a set of reference values determined from a model of an ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values.

16. The method of claim 15, wherein input data X is representative of visual image data.

17. The method of claim 15, further comprising the step of:

further processing the result Z(v) Y(v) using at least one step selected from a group consisting of (i) applying a desired signal processing technique to the result Z(v) Y(v), (ii) providing the result Z(v) Y(v) as a control term to vary amplitude spectrum of the function filter function W(v), (iii) quantizing the result Z(v) Y(v), (iv) quantizing the result Z(v) Y(v) by comparing the result to a set of reference values determined from the result'"'"'s ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values, (iv) quantizing the result Z(v) Y(v) further by comparing the result to a set of reference values determined from a model of an ensemble average probability density function and generating a quantization value for the result corresponding to a member of the set of reference values.

19. The signal processing system of claim 18, wherein input signal X is representative of visual image data.

20. The signal processing system of claim 18, wherein B(v) has at least one characteristic selected from a group consisting of (i) B(v) is proportional to v, (ii) B²(v) is proportional to [a₁+vⁿ]/[a₂+v^m] where a₁ and a₂ are constants and n and m are integers, (iii) B²(v) is represented by vector [0 . . . 0-12-10 . . . 0] where ellipses represent any number of zeros, and (iv) where a>

0 and B²(v) is represented by a matrix;

12. A method of characterizing a data processing operation which processes input data X to form a result Y, comprising the following steps:

forming a function W, where |W| is proportional to |Y|/|X|;

forming a function U, where |U| is equal to

[|W|(1−

|W|)]^½

;

and

applying the function U to the input data X to obtain the result U(v) X(v).

15. A method of characterizing a data processing operation that processes input data X to form a result Y, comprising the following steps:

forming a function W, where |W| is proportional to |Y|/|X|;

forming a function Z, where |Z| is equal to

[(1−

|W|)/|W|]^½

;

and

applying the function Z to the output data Y to obtain the result Z(v) Y(v).

18. A signal processing system, comprising:

a data input node for inputting a signal X to be processed;

a digital computing apparatus including at least a central processor unit and memory programmed to operate on an input signal to implement operations of;

forming a first processing function W, wherein the amplitude spectrum W(v) is given by

W(v)=(1+b²B(v)²)<sup>−

1</sup>;

 

where b is a constant and B(v) is a positive, real valued function of frequency;

forming a second processing function U having an amplitude spectrum given by

[|W|(1−

|W|)]^½

;

applying the second processing function U to the input data;

performing a desired signal processing operation on the result of applying the second processing function U to the input data; and

a display device for displaying a result of operations implemented by said digital computing apparatus.

21. A method of processing a signal, comprising the following steps:

providing a signal generated by a stochastic process, the signal including a stochastic source component and a random process component;

inputting the signal to a filter having an amplitude spectrum W₁(v), where v is a frequency component of the signal, the filter output representing a weighting of the frequency component a greater amount if it is more likely attributable to the stochastic source than to the random process; and

weighting the output of the filter by a function having an amplitude spectrum W₂(v), where W₂(v) is given by

[(1−

W₁(v))/W₁(v)]^½

.

22. The method of claim 21, further comprising the step of:

providing the weighted output of the filter as a control signal to vary the amplitude spectrum of the filter, wherein the control signal has an expectable power spectrum proportional to |W₁(v)W₂(v)|.

23. The method of claim 21, further comprising the step of quantizing an output of the filter so controlled.

24. The method of claim 21, wherein amplitude spectrum W₁(v) is represented by W₁(v)=[1+b²B(v)²]⁻

1, where b is a constant and B(v) is a positive, real valued function of frequency.

25. The method of claim 24, wherein B(v) is proportional to v^±

n, where n is an integer.

27. The signal processor of claim 26, wherein B(v) is proportional to v^±

n, where n is an integer.

28. The signal processor of claim 26, wherein the filter means includes a resistive network having an adjustable parameter which is varied by a control signal.

29. The signal processor of claim 28, wherein the adjustable parameter is a ratio of sheet resistance to shunt resistance.

26. A signal processor, comprising:

means for inputting data X(v) representing a desired signal component plus an undesired contamination component;

filter means for filtering the data so input, the filter means having an amplitude spectrum W(v), where |W(v)|[1+b²B(v)²]<sup>−

1</sup>, B(v) is a positive, real valued function, and b is a positive number;

means for weighting an output of the filter means by U(v), where |U(v)|=[W(v)(1−

W(v))]^½

;

means for providing a weighted output of the filter means as a control signal to vary amplitude spectrum of the filter means; and

means for processing X(v) to form the result U(v) X(v).

30. A method of enhancing regions of an image in which contrast discontinuities are present, the image formed from a plurality of signals generated by a stochastic process, wherein each signal includes a stochastic source component and a random process component, the method comprising the following steps:

weighting the signals by a filter function having an amplitude spectrum W₁(v), where v is a frequency component of the signal, the amplitude spectrum acting to selectively weight the frequency component greater if it is more likely attributable to the stochastic source than to the random process;

weighting the output of the filter function by a function having an amplitude spectrum W2(v), where W₂(v) is given by

[(1−

W1(v))/W₁(v)]^½

;

and

adding a result of weighting output of the filter function by W_{2 (v)} to output of the filter function to form components of an enhanced image.

31. The method of claim 30, wherein amplitude spectrum W₁(v) is given by

W₁(v)=(1+b²B(v)²)⁻

1,

where b is a constant and B(v) is a positive, real valued function of frequency.

32. The method of claim 30, wherein B(v) is proportional to v^±

n, where n is an integer.