Differentiation of illumination and reflection boundaries
View Patent ↗The present invention provides methods and apparatus for image processing in which brightness boundaries of an image are identified and analyzed in at least two, and more preferably three or more, spectral bands to distinguish illumination boundaries from reflectance boundaries. For example, in one embodiment of the invention, a brightness boundary of the image can be identified as an illumination boundary if at least two wavelength bands of the image exhibit a substantially common shift in brightness across the boundary.
1. A method of distinguishing illumination boundaries from reflectance boundaries in an image of a scene, comprising: a computer executing the following steps:
identifying a plurality of brightness boundaries in the image, each brightness boundary separating a high brightness side from a low brightness side, for each of three selected wavelength bands of the image and each of the brightness boundaries, determining image brightness on each side of the boundary, for each of said wavelength bands and each of the brightness boundaries, determining a brightness difference across the boundary, for each of said brightness boundaries, generating a three-component vector wherein each component is formed as a ratio of an absolute value of brightness difference across the boundary relative to brightness on the bright side of the boundary corresponding to one of the wavelength bands, normalizing said vectors, identifying a boundary as an illumination boundary if a normalized vector corresponding to said boundary forms a cluster with one or more vectors corresponding to other boundaries.
2. The method of claim 1 , further characterizing each of said three-component vectors ( V ) as follows
V
_
=
(
D
L
B
L
′
,
D
M
B
M
′
,
D
S
B
S
′
)
wherein L, M, and S denote said three wavelength bands, D L , D M , and D S denote brighness values in the low brightness side of the boundary for each of the three wavelength bands, respectively, and B′ L , B′ M , B′ S are defined, respectively, as
B′ L =B L −D L ;B′ M =B M −D M ;B′ S =B s =D S ,
wherein B L , B M , and B s are brightness values in the high brightness of the boundary for each of the wavelength bands, respectively.
3. The method of claim 2 , further comprising normalizing said vector V to obtain a normalized vector ( V N ) defined as follows:
V
_
N
=
V
_
V
,
wherein ∥V∥ denotes the norm of V .
4. The method of claim 1 , wherein said wavelength bands include long, medium, and short wavelength bands, respectively.
5. An imaging system, comprising an image-capture device for generating a multi-spectral image of a scene, and an image processing module operating on image to identify one or more brightness boundaries therein, and to differentiate each of said brightness boundaries into illumination and reflectance boundaries by, for each brightness boundary, separating a high brightness side from a low brightness side, for each of three selected wavelength bands of the image and each of the brightness boundaries, determining image brightness on each side of the boundary, for each of said wavelength bands and each of the brightness boundaries, determining a brightness difference across the boundary, for each of said brightness boundaries, generating a three-component vector wherein each component is formed as a ratio of an absolute value of brightness difference across the boundary relative to brightness on the bright side of the boundary corresponding to one of the wavelength bands, normalizing said vectors, identifying a boundary as an illumination boundary if a normalized vector corresponding to said boundary forms a cluster with one or more vectors corresponding to other boundaries.
6. The imaging system of claim 5 , wherein said image processing module comprises a storage for storing an image data received from the image-capture device.
7. The imaging system of claim 6 , wherein said image processing module comprises a processor programmed to operate on said image data to identify said brightness boundaries and to classify them into illumination and reflectance boundaries.
8. The imaging system of claim 7 , wherein said processor comprises a module for operating on said image to determine brightness boundaries therein.
9. The imaging system of claim 8 , wherein said processor further comprises another module for utilizing information regarding said brightness boundaries and said image data to classify said boundaries into illumination and reflectance boundaries.