Recovery of hyperspectral data from image
View Patent ↗A method for approximating spectral data, the method comprising using at least one hardware processor for: providing a digital image comprising data in a first set of spectral bands; providing a dictionary comprising (a) signatures in a second set of spectral bands and (b) values in said first set of spectral bands, wherein said values correspond to the signatures, and wherein said first and second sets of spectral bands are different; and approximating, based on the dictionary, data in said second set of spectral bands of said digital image.
1. An apparatus comprising:
an image sensor configured to capture digital image data;
a color filter array (CFA) operatively coupled to said image sensor, said CFA comprising n color filters, wherein the color filters included in said CFA are selected by a process comprising the following steps:
(i) forming a group of random CFA candidates, each comprising n color filters, from a random subset selected from a set comprising a plurality of color filters each having a respective known spectral frequency response function,
(ii) assigning a quality ranking to each of said CFA candidates, based, at least in part, on simulating, using said respective known spectral frequency response functions, the reconstruction of spectral data from image data acquired using said image sensor with said respective CFA candidate used as said CFA,
(iii) iteratively updating said selection of said CFA candidates in said group, wherein each iteration of said group comprises:
(a) a first proportion comprising a subgroup of said CFA candidates having a highest said quality ranking,
(b) a second proportion comprising new CFA candidates randomly formed from color filters included in a pair of said CFA candidates included in a preceding said iteration,
(c) a third proportion comprising new CFA candidates formed by replacing one of said color filters in a CFA candidate included in a preceding said iteration with a randomly-selected color filter from said set, and
(d) a fourth proportion comprising new CFA candidates formed by randomly-selecting color filters from said set,
(iv) assigning said quality ranking to each of said CFA candidates in each of said iterations of said group,
(v) repeating steps (iii) and (iv) until said quality ranking reaches an optimum value, and
(vi) selecting a final CFA candidate from said group as said CFA, based on said optimum value.
2. The apparatus of claim 1 , wherein said first proportion is 10%, said second proportion is 40%, said third proportion is 10%, and said fourth proportion is 40%.
3. The apparatus of claim 1 , wherein said CFA candidates from a said preceding iteration are randomly-selected for participation in said second and third proportions with a probability that is proportional to their respective quality ranking.
4. The apparatus of claim 1 , wherein n is equal to at least 3.
5. The apparatus of claim 1 , wherein said spectral data is hyperspectral data.
6. The apparatus of claim 1 , wherein said CFA comprises color filters corresponding to the red-green-blue (RGB) spectral bands.
7. The apparatus of claim 1 , wherein said quality ranking is determined, at least in part, based on measuring a root mean square error (RMSE) associated with said reconstruction of said spectral data.
8. A method comprising selecting color filters for inclusion in a color filter array (CFA) comprising n color filters and operatively coupled to an image sensor configured to capture a digital image data, by performing the following steps:
(i) forming a group of random CFA candidates, each comprising n color filters, from a random subset selected from a set comprising a plurality of color filters each having a respective known spectral frequency response function,
(ii) assigning a quality ranking to each of said CFA candidates, based, at least in part, on simulating, using said respective known spectral frequency response functions, the reconstruction of spectral data from image data acquired using said image sensor with said respective CFA candidate used as said CFA,
(iii) iteratively updating said selection of said CFA candidates in said group, wherein each iteration of said group comprises:
(a) a first proportion comprising a subgroup of said CFA candidates having a highest said quality ranking,
(b) a second proportion comprising new CFA candidates randomly formed from color filters included in a pair of said CFA candidates included in a preceding said iteration,
(c) a third proportion comprising new CFA candidates formed by replacing one of said color filters in a CFA candidate included in a preceding said iteration with a randomly-selected color filter from said set, and
(d) a fourth proportion comprising new CFA candidates formed by randomly-selecting color filters from said set,
(iv) assigning said quality ranking to each of said CFA candidates in each of said iterations of said group,
(v) repeating steps (iii) and (iv) until said quality ranking reaches an optimum value, and
(vi) selecting a final CFA candidate from said group as said CFA, based on said optimum value.
9. The method of claim 8 , wherein said first proportion is 10%, said second proportion is 40%, said third proportion is 10%, and said fourth proportion is 40%.
10. The method of claim 8 , wherein said CFA candidates from a said preceding iteration are randomly-selected for participation in said second and third proportions with a probability that is proportional to their respective quality ranking.
11. The method of claim 8 , wherein n is equal to at least 3.
12. The method of claim 8 , wherein said spectral data is hyperspectral data.
13. The method of claim 8 , wherein said CFA comprises color filters corresponding to the red-green-blue (RGB) spectral bands.
14. The method of claim 8 , wherein said quality ranking is determined, at least in part, based on measuring a root mean square error (RMSE) associated with said reconstruction of said spectral data.
15. A computer program product comprising a non-transitory computer-readable storage medium having program instructions embodied therewith, the program instructions executable by at least one hardware processor to select color filters for inclusion in a color filter array (CFA) comprising n color filters and operatively coupled to an image sensor configured to capture a digital image data, by performing the following steps:
(i) forming a group of random CFA candidates, each comprising n color filters, from a random subset selected from a set comprising a plurality of color filters each having a respective known spectral frequency response function,
(ii) assigning a quality ranking to each of said CFA candidates, based, at least in part, on simulating, using said respective known spectral frequency response functions, the reconstruction of spectral data from image data acquired using said image sensor with said respective CFA candidate used as said CFA,
(iii) iteratively updating said selection of said CFA candidates in said group, wherein each iteration of said group comprises:
(a) a first proportion comprising a subgroup of said CFA candidates having a highest said quality ranking,
(b) a second proportion comprising new CFA candidates randomly formed from color filters included in a pair of said CFA candidates included in a preceding said iteration,
(c) a third proportion comprising new CFA candidates formed by replacing one of said color filters in a CFA candidate included in a preceding said iteration with a randomly-selected color filter from said set, and
(d) a fourth proportion comprising new CFA candidates formed by randomly-selecting color filters from said set,
(iv) assigning said quality ranking to each of said CFA candidates in each of said iterations of said group,
(v) repeating steps (iv) and (v) until said quality ranking reaches an optimum value, and
(vi) selecting a final CFA candidate from the current said group as said CFA, based on said optimum value.
16. The computer program product of claim 15 , wherein said first proportion is 10%, said second proportion is 40%, said third proportion is 10%, and said fourth proportion is 40%.
17. The computer program product of claim 15 , wherein said CFA candidates from a said preceding iteration are randomly-selected for participation in said second and third proportions with a probability that is proportional to their respective quality ranking.
18. The computer program product of claim 15 , wherein said spectral data is hyperspectral data.
19. The computer program product of claim 15 , wherein said CFA comprises color filters corresponding to the red-green-blue (RGB) spectral bands.
20. The computer program product of claim 15 , wherein said quality ranking is determined, at least in part, based on measuring a root mean square error (RMSE) associated with said reconstruction of said spectral data.