Colour signal processing

• US 20030142222A1
• Filed: 11/27/2002
• Published: 07/31/2003
• Est. Priority Date: 01/12/2000
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: University of East Anglia
• Inventors: Hordley, Stephen
• Granted Patent: US 7,227,586 B2
• Time in Patent Office: Days
• US Class Current: 348/223.1
• CPC Class Codes: G06V 10/56 relating to colour
• CPC Class Codes: H04N 23/88 for colour balance, e.g. white-balance circuits or colour temperature control

Abstract

A method of adjusting the N components of an N component colour signal (such as the 3 components of an RGB colour signal) with respect to colour temperature of illuminating light when imaging a coloured subject, comprising the step of computing the difference between each of the components of the colour signal for different temperatures of illuminating light, using an algorithm based on the spectral sensitivity of the imaging means and Planck'"'"'s equation defining black body illumination to obtain N components whose values are independent of the colour temperature of the illuminating light. Normalisation of the adjusted values renders the values independent of intensity of illumination and can also reduce the processing time for further processing of the signals. A technique for compensating for gamma correction which is commonly built in cameras, is disclosed.

Claims

1. A method of adjusting the N components of an N component colour signal (such as the 3 components of an RGB colour signal) so as to produce component values which are independent of the colour temperature of the illuminating light, comprising the step of computing for each component a new value using the original component value and one other component value and an algorithm incorporating a mathematical operator whose value is related to the spectral sensitivity of the imaging means to different colour temperature illuminants.

3. A method as claimed in claim 1 or 2 wherein the colour signal is derived from a camera, multispectral imaging device, or a scanner, or is a computer generated signal.

4. A method as claimed in any of claims 1 to 3 in which the method is typically applied to the N-RGB etc. values of each pixel making up a multi-pixel image.

5. A corrected RGB signal as claimed in any of claims 1 to 4 comprising three parts namely

6. A corrected RGB signal as claimed in any of claims 1 to 4, in the form of a vector (R*, G*, B*)−

G* v(u, v, w) which can be written in the form (x, O, y), in which the non-zero components x and y are two values that are independent of the colour temperature of the illuminant.

7. A method as claimed in any of claims 1 to 4 wherein the N-RGB etc. values are normalised with respect to one of the N values to remove intensity dependence and reduce the number of values to (N−

1).

8. A method as claim in claim 7 wherein there are three component R G and B in the original signal and as a result of normalisation these are reduced in number to R/G and B/G and colour intensity independence is obtained by converting these two components, one to zero and the other to (B′

−

R′

.b/a) where B′

is log B/G and R′

is log R/G, and the values of a and b are derived from the values u, v, w for the signal source.

10. A method as claimed in any of claims 1 to 9 comprising the step of calibrating the source in advance to determine the (u, v, w etc (or a, b, c etc)) values for the source.

16. A method of image analysis comprising the steps of comparing illumination colour temperature and/or intensity corrected signals obtained by the method of any of claims 1 to 6, 9, 11 or 13 or 14, or summations or other computational combinations of some or all of the intensity corrected signals, with stored values of such signals and/or similar summations or other computational combinations thereof, to permit object recognition.

17. A computer programmed to perform the comparison set forth in claim 16.

18. A method or apparatus as claimed in any of claims 1 to 6, or 9 to 17 in which the source (e.g. camera) includes RGB etc. correction in each of the N channels for the N components to compensate for the expected non-linearity of a display device such as a CRT based display or monitor to which the signal could be supplied, and a linear relationship between the N-component colour signal and the resulting image is obtained by dividing the log camera responses by the average of the log colour responses in a local neighbourhood of the pixel of interest in the image (e.g. in all or some of the immediately adjacent pixels), prior to calculating the invariant co-ordinates ie corrected component values, in accordance with a method as claimed in any of the aforementioned claims.

19. A method or apparatus as claimed in any of claims 1 to 6 or 9 to 17 in which after the invariant co-ordinate(s) are calculated, the value(s) is/are normalised by the average invariant co-ordinate in a local neighbourhood of a pixel of interest (e.g. for all or some of the immediately adjacent pixels) in the image.

25. A method or apparatus as claimed in any of claims 9 to 24 where the source is a camera, a multispectral imaging device, a scanner or a computer programmed to produce a colour signal.

2. A method of processing component signals making up an N component colour signal (such as RGB signals of a 3 component colour signal) defining an image of an illuminated coloured subject to produce modified signal components defining an image of the same subject as if illuminated under light of a different colour temperature, comprising the steps of:

(1) generating the N-(RGB etc.) components of a colour signal corresponding to an illumination of the subject using illumination of colour temperature T₁;

(2) forming the N log values (R′

G′

B′

etc.) of the N-RGB etc. components of the colour signal so obtained; and

(3) computing the N log values (R*, G*, B* etc.) of the N-RGB etc. components of the colour signal for a different illumination colour temperature T₂ using the equation;

(R* G* B* . . . )=(R′

G′

B′

. . . )+X (u, v, w . . . )

where X is proportional to (T1−

T2) and u, v, w etc. are constants that depend only on the spectral sensitivities of the signal source and Planck'"'"'s equation defining black body illumination; and

(4) optionally processing and/or storing and/or controlling a display device using the N-(R*, G*, B* . . . ) values or signals derived therefrom, or converting the log values (R*, G*, B* . . . ) to N appropriate RGB etc. values, for processing and/or display and/or storing.

9. A method of processing the N component signals making up a colour signal (such as RGB signals of a 3 component colour signal) defining an image of an illuminated coloured subject, to produce modified signal components defining an image of the same subject as if illuminated by a different colour temperature (T), comprising the steps of:

(1) generating the N-(RGB etc) components of a colour signal corresponding to an illumination of the subject using illumination of colour temperature T;

(2) normalising the N components of the colour signal with respect to one of the components (e.g. G) to provide (N−

1) co-ordinates (e.g. R/G, B/G in a 3 component RGB signal) so that each of the resulting components is independent of illumination intensity,

(3) computing the log values of these (N−

1) co-ordinates, (such that log(R/G)=R′ and

log (B/G)=B′

in the case of an RGB signal),

(4) adjusting the values of (R′

, B′

, etc.) to produce new values (R*, B*, etc.) corresponding to the values which would apply if the colour temperature of the illumination is T₂ using the equation (R*, B*, etc.)=(R′

, B′

, etc.)+X(a, b, etc.) where X is proportional to (T₁−

T₂) and a and b etc. are constants that depend only on the spectral sensitivities of the source and Planck'"'"'s equation defining black-body illumination; and

(5) optionally converting the values (R*, B*, etc.) to new (RGB etc.) values or using the new values, for processing and/or comparison and/or storage and/or display.

11. A method of processing a colour signal from a colour signal source so as to eliminate dependency on colour temperature and/or intensity variation by using a mathematical algorithm derived from the source sensitivities to different wavelengths of light in a signal processing computer supplied with the colour signal, to provide new (corrected) values for the components making up the colour signal which are independent of the colour temperature of the light illuminating (or deemed to illuminate) the subject described by the colour signal.

12. A computer when programmed so as to perform the method of any of claims 2-6, 9 or 11.

13. A method as claimed in any of claims 2-6, 9 or 11 further comprising the step of converting corrected component signal values as aforesaid into a grey scale or pseudo colour signal for processing and/or comparison and/or storage and/or for forming a grey scale or pseudo colour image in place of a true-colour image in a display.

14. A method as claimed in claim 13 wherein a 1-dimensional colour co-ordinate is expressed as a function of chromaticity, so as to remove dependence on the illumination colour temperature by converting an 2 component colour signal to a single component grey-scale signal or pseudo colour signal, and a comparison is performed by the step of mapping the grey-scale values (or pseudo colour values) to grey-scale values (or pseudo colour values) of a similar scene observed under reference lighting conditions.

15. A method of colour-based object recognition employing colour signals as processed and/or normalised as claimed in any of claims 13 or 14 wherein objects are represented by the distributions of invariant co-ordinates calculated from colour images of the objects obtained by the processing to produce grey-scale histograms (since the invariant coordinates are coded as grey-scale), and recognition is performed by distribution comparison whereby the distribution of an unknown object is compared with known object distributions stored in a database, and the closest match identifies the unknown object.

20. A method of image enhancement when processing a colour signal in a system in which the log of the colour response is composed of three parts, a vector due to the intensity of the illumination, a vector due to the illumination colour temperature, and a vector due to reflective properties of the surface illumiated for the purpose of producing the image, in which the magnitude of the intensity vector, or the magnitude of the illumination colour temperature vector is adjusted for each pixel so as to change the effective illumination for each pixel in a display of the image using the processed signal.

21. A method of enhancement as claimed in claim 20 in which the colour warmness or colour depth is increased or decreased by altering the magnitude of the illumination colour temperature vector of the signal.

22. A method as claimed in claim 21 wherein the magnitude is altered by scaling all illumination colour temperature vectors for all pixels in the image.

23. A method as claimed in claim 21 wherein the magnitude of the illumination intensity vector is increased or decreased for all pixels in an image so as to brighten or darken the overall picture, to allow enhancement of a final image without altering the original illumination.

24. A method as claimed in any of claims 21 to 23 wherein the vector alteration is performed on a frame basis, or on selected regions of a frame made up of many pixels, or on a pixel by pixel basis.