Claims
- 1. A method processing an image of a scene discernible by the human eye comprising the steps of:
- (a) operating a photographic image processing system which transforms a photographic image of a scene discernible by the human eye into electrical signal representative of respective picture regions of a transformed version of said photographic image, said transformed version of said photographic image, when viewed by the human eye, containing discernible noise artifacts resulting from the transformation of said photographic image into said electrical signals; and
- (b) processing said electrical signals to reduce noise artifacts introduced into said transformed version of said photographic image as a result of performing step (a), and thereby produce a resulting processed version of said transformed version of said photographic image which, upon being converted into an output image discernible by the human eye, will not contain such noise artifacts that are discernible to the human eye, said processing step (b) comprising the steps of:
- (b1) selecting a target picture region within said transformed version of said photographic image,
- (b2) prescribing a neighborhood of picture regions about said target picture region,
- (b3) for the neighborhood prescribed in step (b2), generating an electrical signal representative of a goodness of fit (x.sup.2) and an estimated value of the image picture region at said target picture region in accordance with a least squares best fit plane for said prescribed neighborhood;
- (b4) iterating steps (b2) and (b3) for respectively different types and respectively different positions of said prescribed neighborhood;
- (b5) generating an output picture region for said target picture region in accordance with a summation of all estimated values of the image picture region at said target picture region with weights based upon said goodness of fit (X.sup.2), and
- (b6) iterating steps (b1) through (b5) for all picture regions of said transformed version of said photographic image, so as to form said resulting processed version of said transformed version of said photographic image, in which said noise artifacts that have been introduced into said transformed version of said photographic image as a result of performing step (a), are reduced.
- 2. A method according to claim 1, further comprising the step of:
- (c) applying said resulting processed version of said transformed version of said photographic image formed by step (b6) to an output image reproduction device, which generates said output image that does not contain said noise artifacts discernible to the human eye.
- 3. A method according to claim 1, wherein step (b4) includes iterating steps (b2) and (b3) for respectively different shapes and orientations of said prescribed neighborhood.
- 4. A method according to claim 1, wherein in step (b5), rather generating an output picture region for said target picture region in accordance with said summation of all estimated values of the image picture region at said target picture region with weights based upon said goodness of fit, said output image picture region for said target picture region is generated in accordance with a single estimated value associated with the largest weight.
- 5. A method according to claim 1, wherein said goodness of fit (X.sup.2) is a measure of edge detection, texture detection, and image segmentation.
- 6. A method according to claim 1, wherein said target picture region has
- selected neighborhood sizes of:
- 21.times.21, 11.times.11, 5.times.5, 1.times.1,
- empirical weights of:
- W.sub.21.times.21 =64, W.sub.11.times.11 =16, W.sub.5.times.5 =16, W.sub.1.times.1 =1,
- and
- rolloff parameters of:
- .tau..sub.21.times.21 =.tau..sub.11.times.11 =.tau..sub.5.times.5 =.tau..sub.1.times.1 =1.
- 7.
- 7. A method according to claim 1, wherein said goodness of fit is defined as: ##EQU5## and the estimated code value of a target pixel positioned at (i,j) is:
- z.sub.ij.sup.kl,.alpha. =.alpha..sub.kl +b.sub.kl.spsb.i +c.sub.klj
- and the computed noise reduced code value of the target pixel at position (i,j) is: ##EQU6## where the sums are over all kl,.alpha. such that pixel (i,j).epsilon.N.sub.Kl,.alpha. and where the weighting function is: ##EQU7## wherein N.sub.kl,.alpha. is a neighborhood of type .alpha. centered at pixel (k,l), W.sub..alpha. is an empirically chosen weight for each neighborhood type, n.sub..alpha. is the number of pixels in the neighborhood, .nu..sub..alpha. is the degrees of freedom for the least squares fit (.nu..sub..alpha. =n.sub..alpha. -3), r.sub..alpha. is an empirically chosen rolloff parameter of order unity, z.sub.ij is a code value, .sigma..sub.ij is the standard deviation at pixel (i,j), X.sub.kl,.alpha..sup.2 is the goodness of fit for the linear least squares regression of a plane on neighborhood N.sub.kl,.alpha., and z.sub.ij.sup.kl,.alpha. is the estimate of the regression at pixel location (i,j), where pixel (i,j) is some pixel in the neighborhood N.sub.kl,.alpha.., and a,b,c are coefficients which minimize the goodness of fit.
- 8. A method according to claim 1, wherein step (a) comprises operating a digital photographic image-transforming device which transforms said photographic image of a scene discernible by the human eye into digital signals representative of respective digital code values of picture regions of an array of a digitally transformed version of said photographic image of a scene.
- 9. A method processing an image of a scene discernible by the human eye comprising the steps of:
- (a) operating a photographic image processing system which transforms a photographic image of a scene discernible by the human eye into electrical signals representative of respective picture regions of an array of picture regions of a transformed version of said photographic image, said array of picture regions of said transformed version of said photographic image, when viewed by the human eye, containing discernible noise artifacts resulting from the transformation of said photographic image into said electrical signals;
- (b) processing said electrical signals to smooth and reduce noise artifacts introduced into said transformed version of said photographic image as a result of performing step (a), so as to obtain a processed version of said transformed version of said photographic image which, upon being converted into an output image discernible by the human eye, will not contain such noise artifacts that are discernible to the human eye, said processing step (b) comprising the steps of:
- (b1) selecting a target picture region position from said array of picture regions of said transformed version of said photographic image,
- (b2) prescribing a neighborhood of picture regions about said target picture region position;
- (b3) for the neighborhood prescribed in step (b2), generating a signal representative of the least squares best fit plane for said prescribed neighborhood; and
- (b4) generating an electrical signal representative of an output image picture region for said target picture region position in accordance with the least squares best fit signal generated in step (b3); and
- (c) operating an output image reproduction device in accordance with respective electrical signals generated in step (b4), so as to generate said output image that does not contain said noise artifacts discernible to the human eye.
- 10. A method according to claim 9, wherein step (a) comprises operating a digital photographic image-transforming device which transforms said photographic image of a scene discernible by the human eye into digital signals representative of respective digital code values of picture regions of an array of a digitally transformed version of said photographic image of a scene.
Parent Case Info
This is a continuation of application Ser. No. 07/781,468, filed Oct. 23, 1991, U.S. Pat. No. 5,533,149, which is a continuation of U.S. Ser. No. 399,135, filed Aug. 28, 1989, now abandoned.
US Referenced Citations (13)
Non-Patent Literature Citations (1)
Entry |
Panda et al., "Recursive Least Squares Smoothing of Noise in Images," IEEE Transactions on Acoustics, Speech and Signal Processing, vol. ASSP-25, No. 6, Dec. 1977, pp. 520-524. |
Continuations (2)
|
Number |
Date |
Country |
Parent |
781468 |
Oct 1991 |
|
Parent |
399135 |
Aug 1989 |
|