Method for multiframe Wiener restoration of noisy and blurred image sequences

• US 5,550,935 A
• Filed: 04/13/1995
• Issued: 08/27/1996
• Est. Priority Date: 07/01/1991
• Status: Expired due to Fees

• Current Assignee: Eastman Kodak Company
• Original Assignee: Eastman Kodak Company
• Inventors: Erdem, Arif T., Sezan, Muhammed I., Ozkan, Mehmet K.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Johns, Andrew W.
• Application Number:
  US08/421,138
• Time in Patent Office: 502 Days
• Field of Search: 382/103, 382/107, 382/255, 382/260, 382/263, 382/278, 348/241, 348/349, 348/352, 348/607, 348/625
• US Class Current: 382/260
• CPC Class Codes: G06T 7/262 using transform domain methods, e.g. Fourier domain methods
• CPC Class Codes: H04N 19/527 Global motion vector estimation

Abstract

A method for performing multiframe Wiener restoration of noisy and blurred image sequences that provides either a cross-correlated multiframe Wiener restoration or a motion-compensated multiframe Wiener restoration. The cross-correlated multiframe restoration uses interframe correlations explicitly through the estimated cross-power spectra of the frames, and is computationally efficient in its calculation of the Wiener estimate by using either a closed form solution or using a small-size matrix inversion. The motion-compensated multiframe Wiener restoration is used when the interframe motion is a global shift which is either known a priori, or estimated using a motion estimation algorithm. The motion-compensated multiframe Wiener filter has a closed-form solution and does not require any matrix inversion.

Claims

1. A method of generating a sequence of images of a scene, in which at least one of blurring and noise, perceptible by a human observer as a result of relative motion between an image capture device and said scene, is reduced, said method comprising the steps of:

(a) generating a sequence of N image frames of said scene by performing the steps of

(a1) providing an image capture device which, when operated, generates said Sequence of N image frames of said scene in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present, and

(a2) operating said image capture device provided in step (a1), so as to cause said image capture device to generate said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present;

(b) processing said sequence of N image frames of said scene as generated by said image capture device as operated in step (a2) to remove said at least one of blurring and noise present in said sequence of N image frames by performing the steps of;

(b1) digitizing said sequence of N image frames of said scene as generated by said image capture device in step (a2);

(b2) dividing the digitized sequence of images into groups of N frames;

(b3) estimating power spectra and cross power spectra of the N digitized image frames;

(b4) providing as inputs to a cross-correlated multiframe Wiener filter;

the estimation of the power spectra and cross power spectra of the N digitized frame, noise power spectra of the N digitized frames, and blur parameters;

(b5) performing cross-correlated multiframe Wiener filtering using said inputs to obtain N restored image frames as an output; and

(b6) repeating steps (b2)-(b5) for every other group of N frames in the digitized sequence of images to produce a restored image sequence in which said at least one of blurring and noise has been reduced;

(c) coupling the digitized restored image sequence produced in step (b6) to an image reproduction output device; and

(d) operating said image reproduction output device to which said digitized restored image sequence produced in step (b6) is coupled in step (c), and thereby causing said image reproduction output device to reproduce said restored image sequence from which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that a human observer can obtain when viewing said restored image sequence of said scene, as reproduced by said image reproduction output device.

2. The method of claim 1, wherein the noise power spectra of the N digitized images is known a priori.

3. The method of claim 1, wherein the noise power spectra of the N digitized images is calculated from the N digitized images themselves by computing sample variances over uniform regions of the image frames.

4. The method of claim 1, wherein the blur parameters are known a priori.

5. The method of claim 1, wherein the power spectra and cross power spectra of the N digitized image frames in a given group are estimated from a set of sharp prototype image frames.

6. The method of claim 1, wherein the step (b5) includes the step of computing discrete Fourier transform samples of estimates of the N digitized image frames in parallel and computing inverse discrete Fourier transform to obtain image frame estimates in a spatial domain.

7. The method of claim 6, wherein the step of computing discrete Fourier transform samples of estimates of the N digitized image frames in parallel is performed according to the equation:

##EQU42## where F_i,k are elements of two-dimensional discrete Fourier transform of the estimate of the i^th frame, i=1, 2, . . . , N, P<sub>f;

ip</sub>,k are the elements of the diagonalized cross-correlation matrix of the actual i^th and the p^th frames of the N digitized frames, H_p,k^* are the elements of the diagonalized blur matrix for the p^th frame, and Z_pq,k are the elements of an M² ×

M² matrix that are determined by inverting an N×

N matrix, in general, whose elements are defined in terms of the elements of blur, diagonalized cross-correlation, and diagonalized noise autocorrelation matrices, and G_q,k are the elements of two-dimensional discrete Fourier transform of the degraded version of a q^th frame.

8. The method of claim 1, wherein step (b5) includes the step of computing discrete Fourier transform samples of estimates of the N digitized image frames in parallel is performed according to the equation:

##EQU43## where F_i,k are elements of two-dimensional discrete Fourier transform of the estimate of the i^th frame, i=1, 2, . . . , N, P<sub>f;

ip</sub>,k are the elements of the diagonalized cross-correlation matrix of the actual i^th and the p^th frames of the N digitized frames, H_p,k^* are the elements of the diagonalized blur matrix for the p^th frame, and Z_pq,k are the elements of an M² ×

M² matrix that are determined analytically (without matrix inversion) in terms of the elements of blur, diagonalized cross-correlation, and diagonalized noise autocorrelation matrices, and G_q,k are the elements of two-dimensional discrete Fourier transform of the degraded version of a q^th frame.

9. A method according to claim 1, wherein step (a) comprises operating a camera containing an image capture medium, so as to cause said camera to capture successive images of a scene, in which at said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said camera and said scene, may be present, and scanning said image capture medium by means of an image capture medium scanner which converts the images captured on said image capture medium by means of said camera into a sequence of N image frames of said scene.

10. A method of generating a sequence of images of a scene, in which at least one of blurring and noise, perceptible by a human observer as a result of relative motion between an image capture device and said scene, is reduced, said method comprising the steps of:

(a) generating a sequence of N image frames of said scene by performing the steps of

(a1) providing an image capture device which, when operated, generates said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present, and

(a2) operating said image capture device provided in step (a1), so as to cause said image capture device to generate said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present;

(b) processing said sequence of N image frames of said scene as generated by said image capture device in step (a2) to remove said at least one of blurring and noise present in said sequence of N image frames of said scene by performing the steps of;

(b1) digitizing said sequence of N image frames of said scene as generated by said image capture device in step (a2);

(b2) dividing the digitized sequence of N image frames into groups of N frames;

(b3) independently restoring the N digitized frames using single-frame Wiener restoration;

(b4) calculating power spectra and cross power spectra of the N digitized frames of the sequence using the image sequence formed by the resulting images frames as a prototype image sequence;

(b5) providing as inputs to a cross-correlated multiframe Wiener filter;

the estimation of the power spectra and cross power spectra of the N digitized frames, noise power spectra of the N digitized frames, and blur parameters;

(b6) performing cross-correlated multiframe Wiener filtering using said inputs to obtain N restored image frames as an output; and

(b7) repeating steps (b2)-(b6) for every other group of N frames in the digitized sequence of images to produce a restored image sequence;

(c) coupling the digitized restored image sequence of said scene produced in step (b7) to an image reproduction output device; and

(d) operating said image reproduction output device to which said digitized restored image sequence produced in step (b7) is coupled in step (c), and thereby causing said image reproduction output device to reproduce said image sequence of said scene from which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that a human observer can obtain when viewing said restored image sequence of said scene, as reproduced by said image reproduction output device.

11. The method of claim 10, wherein the noise power spectra of the N digitized images is known a priori.

12. The method of claim 10, wherein the noise power spectra of the N digitized images is calculated from the N digitized images themselves by computing sample variances over uniform regions of the image frames.

13. The method of claim 10, wherein the blur parameters are known a priori.

14. The method of claim 10, wherein the step (b6) includes the step of computing discrete Fourier transform samples of estimates of the N digitized image frames in parallel and computing inverse discrete Fourier transform to obtain image frame estimates in a spatial domain.

15. The method of claim 14, wherein the step of computing discrete Fourier transform samples of estimates of the N digitized image frames in parallel is performed according to the equation:

##EQU44## where F_i,k are elements of two-dimensional discrete Fourier transform of the estimate of the i^th frame, i=1, 2, . . . , N, P<sub>f;

ip</sub>,k are the elements of the diagonalized cross-correlation matrix of the actual i^th and the p^th frames of the N digitized frames, H_p,k^* are the elements of the diagonalized blur matrix for the p^th frame, and Z_pq,k are the elements of an M² ×

M² matrix that are determined by inverting an N×

N matrix, in general, whose elements are defined in terms of the elements of blur, diagonalized cross-correlation, and diagonalized noise autocorrelation matrices, and G_q,k are the elements of two-dimensional discrete Fourier transform of the degraded version of a q^th frame.

16. The method of claim 14, wherein step (b5) includes the step of computing discrete Fourier transform samples of estimates of the N digitized image frames in parallel is performed according to the equation:

##EQU45## where F_i,k are elements of two-dimensional discrete Fourier transform of the estimate of the i^th frame, i=1,2, . . . , N, P<sub>f;

ip</sub>,k are the elements of the diagonalized cross-correlation matrix of the actual i^th and the p^th frames of the N digitized frames, H_p,k^* are the elements of the diagonalized blur matrix for the p^th frame, and Z_pq,k are the elements of an M² ×

M² matrix that are determined analytically in terms of the elements of blur, diagonalized cross-correlation, and diagonalized noise autocorrelation matrices, and G_q,k are the elements of two-dimensional discrete Fourier transform of the degraded version of a q^th frame.

17. The method of claim 14, wherein an estimate of the image frames in the spatial domain is obtained according to the equation

18. A method according to claim 10, wherein step (a) comprises operating a camera containing an image capture medium, so as to cause said camera to capture successive images of a scene, in which at said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said camera and said scene, may be present, and scanning said image capture medium by means of an image capture medium scanner which converts the images captured on said image capture medium by means of said camera into a sequence of N image frames of said scene.

19. A method of generating a sequence of images of a scene, in which at least one of blurring and noise, perceptible by a human observer as a result of relative motion between an image capture device and said scene, is reduced, said method comprising the steps of:

(a) generating a sequence of N image frames of said scene by performing the steps of

(a1) providing an image capture device which, when operated, generates a sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present, and

(a2) operating said image capture device provided in step (a1), so as to cause said image capture device to generate said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present;

(b) processing said sequence of N image frames of said scene generated by said image capture device in step (a2) to remove said at least one of blurring and noise present in said sequence of N image frames of said scene by performing the steps of;

(b1) digitizing said sequence of N image frames of said scene generated by said image capture device in step (a2);

(b2) dividing the digitized sequence of N image frames into groups of N neighboring frames for every frame in the sequence of N image frames;

(b3) selecting a frame to be restored from a group of N frames to be used as a reference frame;

(b4) estimating a power spectrum of the actual version of the reference frame using a sharp image other than the actual reference frame as a prototype image;

(b5) estimating a noise power spectrum of the reference frame;

(b6) using a motion estimation algorithm to estimate motion vectors between the reference frame and the other frames in the group of N frames, said motion vectors carrying interframe correlation information;

(b7) providing as inputs to a motion-compensated multiframe wiener filter;

the estimation of the power spectrum of the actual version of the reference frame, the estimation of motion vectors estimation noise power spectrum of the reference frame and blur parameters;

(b8) performing a motion compensated multiframe wiener filtering using said inputs to obtain a restored reference frame as an output; and

(b9) repeating steps (b2)-(b8) for every other frame in the sequence to restore the entire image sequence;

(c) coupling the restored image sequence produced in step (b9) to an image reproduction output device; and

(d) operating said image reproduction output device to which said digitized restored image sequence of said scene produced in step (b9) is coupled in step (c), and thereby causing said image reproduction output device to reproduce said image sequence of said scene from which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that a human observer can obtain when viewing said restored image sequence of said scene as reproduced by said image reproduction output device.

20. The method of claim 19, wherein step (b9) includes determining the quantities P_f,k and P_v,k from the estimate of the power spectrum of the actual version of the reference frame and the estimate of the noise power spectrum of the reference frame, where the quantity P_f,k is the k^th element of the diagonalized cross-correlation matrix of the actual reference frame, and the quantity P_v,k is the k^th element of the diagonalized autocorrelation matrix of the noise of the reference frame.

21. The method of claim 20, wherein step (b9) further includes determining discrete Fourier transform samples of the blur parameters and the sequence of images.

22. The method of claim 21, wherein step (b9) further includes using the motion vector estimates to determine S_i and Φ

_i, where S_i is the matrix representing the integer-valued component of the shift of the i^th frame with respect to the reference frame, and Φ

_i is the matrix representing the residual (non-integer) component of the shift of the i^th frame with respect to the reference frame.

23. The method of claim 22, wherein step (b9) further includes determining C_i,k and Ψ

_i,k according to the following equations;

##EQU46## where C_i,k is the k^th element of the diagonalized S_i matrix, and Ψ

_i,k is the k^th element of the diagonalized Φ

_i matrix.

24. The method of claim 23, wherein step (b9) further includes deteremining a discrete Fourier transform of the estimate of the reference frame according to the equation ##EQU47## where C_j,k is the k^th element of the diagonalized S_j matrix, and Ψ

_j,k is the k^th element of the diagonalized Φ

_j matrix, H_j,k is the k^th element of the diagonalized blur matrix, P<sub>f;

k</sub> is the k^th element of the diagonalized autocorrelation matrix of the actual reference frame, P<sub>v;

k</sub> is the k^th element of the diagonalized noise autocorrelation matrix, and G_j,k is the k^th element of the two-dimensional discrete Fourier transform of the degraded version of the j^th frame.

25. The method of claim 24, wherein step (b9) further includes obtaining an estimate of the reference frame in the spatial domain according to the equation

26. A method according to claim 19, wherein step (a) comprises operating a camera containing an image capture medium, so as to cause said camera to capture successive images of a scene, in which at said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said camera and said scene, may be present, and scanning said image capture medium by means of an image capture medium scanner which converts the images captured on said image capture medium by means of said camera into a sequence of N image frames of said scene.

27. A method of generating a sequence of images of a scene, in which at least one of blurring and noise, perceptible by a human observer as a result of relative motion between a high speed video camera and said scene, is reduced, said method comprising the steps of:

(a) generating a sequence of N image frames of said scene by performing the steps of

(a1) providing a high speed video camera which, when operated, generates said sequence of N image frames of said scene in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said high speed video camera and said scene, may be present, and

(a2) operating said high speed video camera such that said high speed video camera generates said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said high speed video camera and said scene, may be present;

(b) processing said sequence of N image frames of said scene as generated by said high speed video camera as operated in step (a2) to remove said at least one of blurring and noise present in said sequence of N image frames by performing the steps of;

(b1) digitizing said sequence of N image frames of said scene as generated by said high speed video camera in step (a2);

(b2) dividing the digitized sequence of images into groups of N frames;

(b3) estimating power spectra and cross power spectra of the N digitized image frames;

(b4) providing as inputs to a cross-correlated multiframe Wiener filter;

the estimation of the power spectra and cross power spectra of the N digitized frame, noise power spectra of the N digitized frames, and blur parameters;

(b5) performing cross-correlated multiframe Wiener filtering using said inputs to obtain N restored image frames as an output; and

(b6) repeating steps (b2)-(b5) for every other group of N frames in the digitized sequence of images to produce a restored image sequence in which said at least one of blurring and noise has been reduced;

(c) providing an image reproduction device which, when operated, is operative to reproduce a digitized restored image sequence of a scene supplied thereto for viewing by said human observer;

(d) coupling said digitized restored image sequence produced in step (b6) to said image reproduction output device provided in step (c); and

(e) operating said image reproduction output device provided in step (c) and to which said digitized restored image sequence produced in step (b6) is coupled in step (d), and thereby causing said image reproduction output device to reproduce said restored image sequence of said scene from which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that a human observer can obtain when viewing said restored image sequence of said scene, as reproduced by said image reproduction output device.

28. A method according to claim 27, wherein said image reproduction output device comprises a cathode ray tube device, and wherein step (e) comprises operating said cathode ray tube device to which said digitized restored image sequence produced in step (b6) is coupled in step (d), and thereby causing said cathode ray tube device to display said restored image sequence of said scene from which said at least one of blurring and noise has been reduced to said human observer, thereby increasing the amount of information that said human observer obtains when viewing said restored image sequence of said scene, as displayed by said cathode ray tube device.

29. A method according to claim 27, wherein said image reproduction output device comprises a device which records said restored image sequence of said scene on a film medium for viewing by said human observer, and wherein step (e) comprises operating said device which records said restored image sequence of said scene on a film medium for viewing by said human observer to which said digitized restored image sequence produced in step (b6) is coupled in step (d), and thereby causing said device to record said restored image sequence of said scene on a film medium for viewing by said human observer from which said at least one of blurring and noise has been reduced to said human observer, thereby increasing the amount of information that said human observer obtains when viewing said film medium containing restored image sequence of said scene.

30. A method of generating a sequence of images of a scene, in which at least one of blurring and noise, perceptible by a human observer as a result of relative motion between a high speed video camera and said scene, is reduced, said method comprising the steps of:

(a) generating a sequence of N image frames of said scene by performing the steps of

(a1) providing a high speed video camera which, when operated, generates said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said high speed video camera and said scene, may be present, and

(a2) operating said high speed video camera provided in step (a1), so as to cause said high speed video camera to generate said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said high speed video camera and said scene, may be present;

(b) processing said sequence of N image frames of said scene as generated by said high speed video camera in step (a2) to remove said at least one of blurring and noise present in said sequence of N image frames of said scene by performing the steps of;

(b1) digitizing said sequence of N image frames of said scene as generated by said high speed video camera in step (a2);

(b2) dividing the digitized sequence of N image frames into groups of N frames;

(b3) independently restoring the N digitized frames using single-frame Wiener restoration;

(b4) calculating power spectra and cross power spectra of the N digitized frames of the sequence using the image sequence formed by the resulting images frames as a prototype image sequence;

(b5) providing as inputs to a cross-correlated multiframe Wiener filter;

the estimation of the power spectra and cross power spectra of the N digitized frames, noise power spectra of the N digitized frames, and blur parameters;

(b6) performing cross-correlated multiframe Wiener filtering using said inputs to obtain N restored image frames as an output; and

(b7) repeating steps (b2)-(b6) for every other group of N frames in the digitized sequence of images to produce a restored image sequence;

(c) providing an image reproduction device which, when operated, is operative to reproduce a digitized restored image sequence of a scene supplied thereto for viewing by said human observer;

(d) coupling the digitized restored image sequence of said scene produced in step (b7) to said image reproduction output device provided in step (c); and

(e) operating said image reproduction output device to which said digitized restored image sequence produced in step (b7) is coupled in step (d), and thereby causing said image reproduction output device to reproduce said image sequence of said scene from which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that a human observer can obtain when viewing said restored image sequence of said scene, as reproduced by said image reproduction output device.

31. A method according to claim 30, wherein said image reproduction output device comprises a cathode ray tube device, and wherein step (e) comprises operating said cathode ray tube device to which said digitized restored image sequence produced in step (b6) is coupled in step (d), and thereby causing said cathode ray tube device to display said restored image sequence of said scene from which said at least one of blurring and noise has been reduced to said human observer, thereby increasing the amount of information that said human observer obtains when viewing said restored image sequence of said scene, as displayed by said cathode ray tube device.

32. A method according to claim 30, wherein said image reproduction output device comprises a device which records said restored image sequence of said scene on a film medium for viewing by said human observer, and wherein step (e) comprises operating said device which records said restored image sequence of said scene on a film medium for viewing by said human observer to which said digitized restored image sequence produced in step (b6) is coupled in step (d), and thereby causing said device to record said restored image sequence of said scene on a film medium for viewing by said human observer from which said at least one of blurring and noise has been reduced to said human observer, thereby increasing the amount of information that said human observer obtains when viewing said film medium containing restored image sequence of said scene.

33. A method of generating a sequence of images of a scene, in which at least one of blurring and noise, perceptible by a human observer as a result of relative motion between a high speed video camera and said scene, is reduced, said method comprising the steps of:

(a) generating a sequence of N image frames of said scene by performing the steps of

(a1) providing a high speed video camera which, when operated, generates a sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said high speed video camera and said scene, may be present, and

(a2) operating said high speed video camera provided in step (a1), so as to cause said high speed video camera to generate said sequence of N image frames of said scene, in which said at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said high speed video camera and said scene, may be present;

(b) processing said sequence of N image frames of said scene generated by said high speed video camera in step (a2) to remove said at least one of blurring and noise present in said sequence of N image frames of said scene by performing the steps of;

(b1) digitizing said sequence of N image frames of said scene generated by said high speed video camera in step (a2);

(b2) dividing the digitized sequence of N image frames into groups of N neighboring frames for every frame in the sequence of N image frames;

(b3) selecting a frame to be restored from a group of N frames to be used as a reference frame;

(b4) estimating a power spectrum of the actual version of the reference frame using a sharp image other than the actual reference frame as a prototype image;

(b5) estimating a noise power spectrum of the reference frame;

(b6) using a motion estimation algorithm to estimate motion vectors between the reference frame and the other frames in the group of N frames, said motion vectors carrying interframe correlation information;

(b7) providing as inputs to a motion-compensated multiframe Wiener filter;

the estimation of the power spectrum of the actual version of the reference frame, the estimation of motion vectors estimation noise power spectrum of the reference frame and blur parameters;

(b8) performing a motion compensated multiframe Wiener filtering using said inputs to obtain a restored reference frame as an output; and

(b9) repeating steps (b2)-(b8) for every other frame in the sequence to restore the entire image sequence;

(c) providing an image reproduction device which, when operated, is operative to reproduce a digitized restored image sequence of a scene supplied thereto for viewing by said human observer;

(d) coupling the restored image sequence produced in step (b9) to said image reproduction output device provided in step (c); and

(e) operating said image reproduction output device to which said digitized restored image sequence of said scene produced in step (b9) is coupled in step (d), and thereby causing said image reproduction output device to reproduce said image sequence of said scene from which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that a human observer can obtain when viewing said restored image sequence of said scene as reproduced by said image reproduction output device.

34. A method according to claim 33, wherein said image reproduction output device comprises a cathode ray tube device, and wherein step (e) comprises operating said cathode ray tube device to which said digitized restored image sequence produced in step (b6) is coupled in step (d), and thereby causing said cathode ray tube device to display said restored image sequence of said scene from which said at least one of blurring and noise has been reduced to said human observer, thereby increasing the amount of information that said human observer obtains when viewing said restored image sequence of said scene, as displayed by said cathode ray tube device.

35. A method according to claim 33, wherein said image reproduction output device comprises a device which records said restored image sequence of said scene on a film medium for viewing by said human observer, and wherein step (e) comprises operating said device which records said restored image sequence of said scene on a film medium for viewing by said human observer to which said digitized restored image sequence produced in step (b6) is coupled in step (d), and thereby causing said device to record said restored image sequence of said scene on a film medium for viewing by said human observer from which said at least one of blurring and noise has been reduced to said human observer, thereby increasing the amount of information that said human observer obtains when viewing said film medium containing restored image sequence of said scene.

36. An apparatus for generating a sequence of images of a scene, in which at least one of blurring and noise perceptible by a human observer is reduced, said apparatus comprising:

an image capture device which is operative to view a scene and generate a sequence of N image frames of said scene, wherein at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present in said sequence of N image frames;

a digital image processor which is programmed to process said sequence of N image frames of said scene as generated by said image capture device, in order to remove said at least one of blurring and noise present in said sequence of N image frames, by performing the steps of;

(1) digitizing said sequence of N image frames of said scene as generated by said image capture device;

(2) dividing the digitized sequence of images into groups of N frames;

(3) estimating power spectra and cross power spectra of the N digitized image frames;

(4) providing as inputs to a cross-correlated multiframe Wiener filter;

the estimation of the power spectra and cross power spectra of the N digitized frame, noise power spectra of the N digitized frames, and blur parameters;

(5) performing cross-correlated multiframe Wiener filtering using said inputs to obtain N restored image frames as an output; and

(6) repeating steps (2)-(5) for every other group of N frames in the digitized sequence of images to produce a restored image sequence in which said at least one of blurring and noise has been reduced; and

an image reproduction device, which is coupled to receive said digitized restored image sequence produced by said digital image processor and is operative to reproduce, for viewing by a human observer, a digitized restored image sequence of said scene in which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that said human observer can obtain when viewing said restored image sequence of said scene, as reproduced by said image reproduction output device.

37. An apparatus according to claim 36, wherein said image capture device comprises a high speed video camera.

38. An apparatus according to claim 36, wherein said image reproduction output device comprises a cathode ray tube device, which is operative to display said restored image sequence of said scene from which said at least one of blurring and noise has been reduced by said digital image processor.

39. An apparatus according to claim 36, wherein said image reproduction output device comprises a device which records said restored image sequence of said scene on a film medium for viewing by said human observer.

40. An apparatus for generating a sequence of images of a scene, in which at least one of blurring and noise perceptible by a human observer is reduced, said apparatus comprising:

an image capture device which is operative to view a scene and generate a sequence of N image frames of said scene, wherein at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present in said sequence of N image frames;

a digital image processor which is programmed to process said sequence of N image frames of said scene as generated by said image capture device, in order to remove said at least one of blurring and noise present in said sequence of N image frames, by performing the steps of;

(1) digitizing said sequence of N image frames of said scene as generated by said image capture device;

(2) dividing the digitized sequence of N image frames into groups of N frames;

(3) independently restoring the N digitized frames using single-frame Wiener restoration;

(4) calculating power spectra and cross power spectra of the N digitized frames of the sequence using the image sequence formed by the resulting images frames as a prototype image sequence;

(5) providing as inputs to a cross-correlated multiframe Wiener filter;

the estimation of the power spectra and cross power spectra of the N digitized frames, noise power spectra of the N digitized frames, and blur parameters;

(6) performing cross-Correlated multiframe Wiener filtering using said inputs to obtain N restored image frames as an output; and

(7) repeating steps (2)-(6) for every other group of N frames in the digitized sequence of images to produce a restored image sequence;

an image reproduction device, which is coupled to receive said digitized restored image sequence produced by said digital image processor and is operative to reproduce, for viewing by a human observer, a digitized restored image sequence of said scene in which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that said human observer can obtain when viewing said restored image sequence of said scene, as reproduced by said image reproduction output device.

41. An apparatus according to claim 40, wherein said image capture device comprises a high speed video camera.

42. An apparatus according to claim 40, wherein said image reproduction output device comprises a cathode ray tube device, which is operative to display said restored image sequence of said scene from which said at least one of blurring and noise has been reduced by said digital image processor.

43. An apparatus according to claim 40, wherein said image reproduction output device comprises a device which records said restored image sequence of said scene on a film medium for viewing by said human observer.

44. An apparatus for generating a sequence of images of a scene, in which at least one of blurring and noise perceptible by a human observer is reduced, said apparatus comprising:

an image capture device which is operative to view a scene and generate a sequence of N image frames of said scene, wherein at least one of blurring and noise, perceptible by a human observer as a result of relative motion between said image capture device and said scene, may be present in said sequence of N image frames;

a digital image processor which is programmed to process said sequence of N image frames of said scene as generated by said image capture device, in order to remove said at least one of blurring and noise present in said sequence of N image frames, by performing the steps of;

(1) digitizing said sequence of N image frames of said scene generated by said image capture device;

(2) dividing the digitized sequence of N image frames into groups of N neighboring frames for every frame in the sequence of N image frames;

(3) selecting a frame to be restored from a group of N frames to be used as a reference frame;

(4) estimating a power spectrum of the actual version of the reference frame using a sharp image other than the actual reference frame as a prototype image;

(5) estimating a noise power spectrum of the reference frame;

(6) using a motion estimation algorithm to estimate motion vectors between the reference frame and the other frames in the group of N frames, said motion vectors carrying interframe correlation information;

(7) providing as inputs to a motion-compensated multiframe Wiener filter;

the estimation of the power spectrum of the actual version of the reference frame, the estimation of motion vectors estimation noise power spectrum of the reference frame and blur parameters;

(8) performing a motion compensated-multiframe Wiener filtering using said inputs to obtain a restored reference frame as an output; and

(9) repeating steps (2)-(8) for every other frame in the sequence to restore the entire image sequence; and

an image reproduction device, which is coupled to receive said digitized restored image sequence produced by said digital image processor and is operative to reproduce, for viewing by a human observer, a digitized restored image sequence of said scene in which said at least one of blurring and noise has been reduced, thereby increasing the amount of information that said human observer can obtain when viewing said restored image sequence of said scene, as reproduced by said image reproduction output device.

45. An apparatus according to claim 44, wherein said image capture device comprises a high speed video camera.

46. An apparatus according to claim 44, wherein said image reproduction output device comprises a cathode ray tube device, which is operative to display said restored image sequence of said scene from which said at least one of blurring and noise has been reduced by said digital image processor.

47. An apparatus according to claim 44, wherein said image reproduction output device comprises a device which records said restored image sequence of said scene on a film medium for viewing by said human observer.