Method of improving a digital image as a function of its dynamic range

• US 6,834,125 B2
• Filed: 06/25/2001
• Issued: 12/21/2004
• Est. Priority Date: 06/25/2001
• Status: Expired due to Fees

• Current Assignee: National Aeronautics and Space Administration (Government of the United States of America), Scientific Technologies Incorporated
• Original Assignee: Scientific Technologies Incorporated, United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration
• Inventors: Jobson, Daniel J., Woodell, Glenn A., Rahman, Zia-ur
• Primary Examiner(s): Couso, Von J.
• Assistant Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US09/888,701
• Publication Number:
  US 20030072496A1
• Time in Patent Office: 1,275 Days
• Field of Search: 382/254, 382/260, 382/274, 382/162, 382/279, 382/191, 358/447, 358/448, 358/461, 358/530, 358/532
• US Class Current: 382/274
• CPC Class Codes: G06T 5/20 using local operators

Abstract

The present invention is a method of processing a digital image that is initially represented by digital data indexed to represent positions on a display. The digital data is indicative of an intensity value I_i(x,y) for each position (x,y) in each i-th spectral band. A classification of the image based on its dynamic range is then defined in each of the image'"'"'s S spectral bands. The intensity value for each position in each i-th spectral band is adjusted to generate an adjusted intensity value for each position in each i-th spectral band in accordance with $\sum _{n=1}^NW_n\left(\log I_i\left(x,y\right)-\log \left[I_i\left(x,y\right)*F_n\left(x,y\right)\right]\right),i=1,\dots ,S$

where W_n is a weighting factor, “*” is the convolution operator and S is the total number of unique spectral bands. For each n, the function F_n(x,y) is a unique surround function applied to each position (x,y) and N is the total number of unique surround functions. Each unique surround function is scaled to improve some aspect of the digital image, e.g., dynamic range compression, color constancy, and lightness rendition. The adjusted intensity value for each position in each i-th spectral band of the image is then filtered with a filter function that is based on the dynamic range classification of the image.

Claims

1. A method of processing a digital image, comprising the steps of:

providing digital data indexed to represent positions of an image having S spectral bands for simultaneous output on a display, said digital data being indicative of an intensity value I_i(x,y) for each position (x,y) in each i-th spectral band;

evaluating features of said image indicative of dynamic range of said image in each of said S spectral bands to thereby identify a class associated with said dynamic range;

adjusting said intensity value for said each position in each i-th spectral band to generate an adjusted intensity value for said each position in each i-th spectral band in accordance with $\underset{n=1}{\overset{N}{\sum <br><br>}}<br><br>W_n<br><br>\left(\log <br><br>\text{\hspace{1em}<br><br>}<br><br>I_i<br><br>\left(x,y\right)-\log <br><br>\left[I_i<br><br>\left(x,y\right)*F_n<br><br>\left(x,y\right)\right]\right),i=1,\mathrm{\dots <br><br>}<br><br>\text{\hspace{1em}<br><br>},S$

where S is the number of unique spectral bands included in said digital data and, for each n, W_n is a weighting factor and F_n(x,y) is a unique surround function applied to said each position (x,y) and N is the total number of unique surround functions;

selecting a filter function based on said class, said filter function so-selected being optimized in terms of offset and gain for said dynamic range associated with said class; and

filtering said adjusted intensity value for said each position of said image in each of said S spectral bands using said filter function so-selected, wherein a filtered intensity value R_i(x,y) is defined.

2. A method according to claim 1 wherein each said unique surround function is a Gaussian function.

3. A method according to claim 2 wherein said Gaussian function is of the form $e^{\frac{-r^2}{c_n^2}}$

satisfying the relationship $k_n\; <li></li>\int <br><br>\int <br><br>e^{\frac{-r^2}{c_n^2}}<br><br><br><br>x<br><br><br><br>y=1$

where

4. A method according to claim 1 further comprising the step of multiplying said filtered intensity value R_i(x,y) by $\log [\frac{B}{}$

 



Ii

(x,y)∑

i=1S

Ii

(x,y)]

to define a color-restored intensity value R′

_i(x,y), where B is a constant.

5. A method according to claim 1 wherein said each position (x,y) defines a pixel of said display.

6. A method according to claim 1 wherein, for each n, W_n=1/N.

7. A method according to claim 1 wherein said features of said image comprises image statistics associated with said image in each of said S spectral bands.

8. A method according to claim 7 wherein said image statistics include brightness and contrast of said image in each of said S spectral bands.

9. A method according to claim 1 further comprising the steps of:

selecting a maximum intensity value V_i(x,y) from the group consisting of said intensity value I_i(x,y) and said filtered intensity value R_i(x,y); and

displaying an improved image using said maximum intensity value V_i(x,y).

10. A method according to claim 4 further comprising the steps of:

selecting a maximum intensity value V_i(x,y) from the group consisting of said intensity value I_i(x,y) and said color-restored intensity value R′

_i(x,y); and

displaying an improved image using said maximum intensity value V_i(x,y).

11. A method of processing a digital image, comprising the steps of:

providing digital data indexed to represent the positions of a plurality of pixels of a J-row by K-column display, said digital data being indicative of an intensity value I(x,y) for each of said plurality of pixels where x is an index of a position in the J-th row of said display and y is an index of a position in the K-th column of said display wherein a J×

K image is defined;

evaluating features of said J×

K image indicative of dynamic range of said J×

K image to thereby identify a class associated with said dynamic range;

convolving said digital data associated with each of said plurality of pixels with a function $e^{\frac{-r^2}{c^2}}$

to form a discrete convolution value for each of said plurality of pixels, said function satisfying the relationship $k<br><br>\int <br><br>\int <br><br>e^{\frac{-r^2}{c^2}}<br><br><br><br>x<br><br><br><br>y=1$

where

12. A method according to claim 11 wherein the value of said constant c is selected to be in the range of approximately 0.01 to approximately 0.5 of the larger of J and K.

13. A method according to claim 11 further comprising the steps of:

selecting, for each of said plurality of pixels, a maximum intensity value V(x,y) from the group consisting of said intensity value I(x,y) and said filtered intensity value R(x,y); and

displaying an improved image using said maximum intensity value V(x,y).

14. A method of processing a digital image, comprising the steps of:

providing digital data indexed to represent the positions of a plurality of pixels of an J-row by K-column display, said digital data being indicative of an intensity value I_i(x,y) for each i-th spectral band of S spectral bands for each of said plurality of pixels where x is an index of a position in the J-th row of said display and y is an index of a position in the K-th column of said display wherein a (J×

K)_i image is defined for each of said S spectral bands and a J×

K image is defined across all of said S spectral bands;

evaluating features of each said (J×

K)_i image indicative of dynamic range of each said (J×

K)_i image to thereby identify a class associated with said dynamic range;

convolving said digital data associated with each of said plurality of pixels in each i-th spectral band with a function $e^{\frac{-r^2}{c_n^2}}$

for n=2 to N to form N convolution values for each of said plurality of pixels in each said i-th spectral band, said function satisfying the relationship $k_n<br><br>\int <br><br>\int <br><br>e^{\frac{-r^2}{c_n^2}}<br><br><br><br>x<br><br><br><br>y=1$

where

 

r=√

{square root over (x²+y²)}

and, for each n, k_n is a normalization constant and c_n is a unique constant;

converting, for each of said plurality of pixels in each said i-th spectral band, each of said N convolution values into the logarithm domain;

converting, for each of said plurality of pixels in each said i-th spectral band, said intensity value into the logarithm domain;

subtracting, for each of said plurality of pixels in each said i-th spectral band, each of said N convolution values so-converted into the logarithm domain from said intensity value so-converted into the logarithm domain, wherein an adjusted intensity value is generated for each of said plurality of pixels in each said i-th spectral band based on each of said N convolution values;

forming a weighted sum for each of said plurality of pixels in each said i-th spectral band using said adjusted intensity values;

selecting a filter function based on said class, said filter function so-selected being optimized in terms of offset and gain for said dynamic range associated with said class; and

filtering said weighted sum for each of said plurality of pixels in each said i-th spectral band with said filter function so-selected, wherein a filtered intensity value R_i(x,y) is defined.

15. A method according to claim 14 wherein the value for each said unique constant c_n is selected to be in the range of approximately 0.01 to approximately 0.5 of the larger of J and K.

16. A method according to claim 14 further comprising the step of multiplying said filtered intensity value R_i(x,y) by $\log [\frac{B}{}$

 



Ii

(x,y)∑

i=1S

Ii

(x,y)]

to define a color-restored intensity value R′

_i(x,y), where B is a constant and S is a whole number greater than or equal to 2.

17. A method according to claim 14 wherein said features of said (J×

K_i image statistics associated with each said (J×

K)_i image.

18. A method according to claim 17 wherein said image statistics include brightness and contrast of each said (J×

K)_i image.

19. A method according to claim 14 further comprising the steps of:

selecting a maximum intensity value V_i(x,y) from the group consisting of said intensity value I_i(x,y) and said filtered intensity value R_i(x,y); and

displaying an improved image using said maximum intensity value V_i(x,y).

20. A method according to claim 16 further comprising the steps of:

selecting a maximum intensity value V_i(x,y) from the group consisting of said intensity value I_i(x,y) and said color-restored intensity value R′

_i(x,y); and

displaying an improved image using said maximum intensity value V_i(x,y).

21. A method of processing a digital image, comprising the steps of:

providing digital data indexed to represent positions of an image having S spectral bands for simultaneous output on a display, said digital data being indicative of an intensity value I_i(x,y) for each position (x,y) in each i-th spectral band;

evaluating features of said image indicative of dynamic range of said image in each of said S spectral bands to thereby identify a class associated with said dynamic range;

adjusting said intensity value for said each position in each i-th spectral band to generate an adjusted intensity value for said each position in each i-th spectral band in accordance with $\underset{n=1}{\overset{N}{\sum <br><br>}}<br><br>W_n<br><br>\left(\log <br><br>\text{\hspace{1em}<br><br>}<br><br>I_i<br><br>\left(x,y\right)-\log <br><br>\left[I_i<br><br>\left(x,y\right)*F_n<br><br>\left(x,y\right)\right]\right),i=1,\mathrm{\dots <br><br>}<br><br>\text{\hspace{1em}<br><br>},S$

where S is a whole number greater than or equal to 2 and defines the total number of spectral bands included in said digital data and, for each n, W_n is a weighting factor and F_n(x,y) is a unique surround function of the form $e^{\frac{-r^2}{c_n^2}}$

satisfying the relationship $k_n<br><br>\int <br><br>\int <br><br>e^{\frac{-r^2}{c_n^2}}<br><br><br><br>x<br><br><br><br>y=1$

where

22. A method according to claim 21 wherein, for each n, W_n=1/N.

23. A method according to claim 21 wherein the value for each said unique constant c_n is selected to be in the range of approximately 0.01 to approximately 0.5 of the larger of J and K.

24. A method according to claim 21 wherein said features of said image comprises image statistics associated with said image in each of said S spectral bands.

25. A method according to claim 24 wherein said image statistics include brightness and contrast of said image in each of said S spectral bands.

26. A method according to claim 21 further comprising the steps of:

selecting a maximum intensity value V_i(x,y) from the group consisting of said intensity value I_i(x,y) and said color-restored intensity value R′

_i(x,y); and

displaying an improved image using said maximum intensity value V_i(x,y).