Method and system for learning a same-material constraint in an image

1. An automated, computerized method for processing an image, comprising the steps of:

providing an image file depicting an image, in a computer memory;

assembling a feature vector for the image file, the feature vector containing information regarding a likelihood that a selected pair of regions of the image file are of a same intrinsic characteristic;

providing a classifier derived from a computer learning technique;

computing a classification score for the selected pair of regions of the image file, as a function of the feature vector and the classifier;

classifying the regions as being of the same intrinsic characteristic, as a function of the classification score;

when a classification score indicates regions as being of the same intrinsic characteristic, defining a constraint between the regions based upon a spatio-spectral relationship defined by the same intrinsic characteristic; and

performing an optimization operation as a function of the constraint to utilize the spatio-spectral relationship defined by the same intrinsic characteristic to generate an intrinsic image corresponding to the image, the intrinsic image comprising a representation of one of material or illumination, expressed as a separate, multi-band representation for the one of material or illumination, independent of the other of the material or illumination, wherein each band corresponds to a segment of the electro-magnetic spectrum, such that when the intrinsic image comprises an illumination image, the illumination image captures the intensity and color of light incident upon each point on the surfaces depicted in the image, and when the intrinsic image comprises a material image, the material image captures reflectance properties of surfaces depicted in the image, including the percentage of each wavelength of light a surface reflects, to segregate the image into an intrinsic component.

2. The method of claim 1 wherein the regions comprise adjacent Type C tokens.

3. The method of claim 1 wherein the same intrinsic characteristic comprises a same material.

4. The method of claim 1 wherein the step of providing a classifier derived from a computer learning technique is carried out by training the classifier using an image database including coded ground truth versions of images in the image database.

5. The method of claim 4 wherein the coded ground truth versions of the images comprise manually coded images to indicate the same intrinsic characteristic.

6. The method of claim 1 wherein the classifier comprises a classifier arranged to process regions of the image that are adjacent and a classifier arranged to process regions of the image that are non-adjacent.

7. A computer system which comprises:

a CPU; and

a memory storing an image file containing an image;

the CPU arranged and configured to execute a routine to assemble a feature vector for the image file, the feature vector containing information regarding a likelihood that a selected pair of regions of the image file are of a same intrinsic characteristic, provide a classifier derived from a computer learning technique, compute a classification score for the selected pair of regions of the image file, as a function of the feature vector and the classifier and classify the regions as being of the same intrinsic characteristic, as a function of the classification score, and wherein operation of the CPU, when a classification score indicates regions as being of the same intrinsic characteristic, includes operation to define a constraint between the regions based upon a spatio-spectral relationship defined by the same intrinsic characteristic, and perform an optimization operation as a function of the constraint to utilize the spatio-spectral relationship defined by the same intrinsic characteristic to generate an intrinsic image corresponding to the image, the intrinsic image comprising a representation of one of material or illumination, expressed as a separate, multi-band representation for the one of material or illumination, independent of the other of the material or illumination, wherein each band corresponds to a segment of the electro-magnetic spectrum, such that when the intrinsic image comprises an illumination image, the illumination image captures the intensity and color of light incident upon each point on the surfaces depicted in the image, and when the intrinsic image comprises a material image, the material image captures reflectance properties of surfaces depicted in the image, including the percentage of each wavelength of light a surface reflects, to segregate the image into an intrinsic component.

8. A computer program product, disposed on a non-transitory computer readable media, the product including computer executable process steps operable to control a computer to: assemble a feature vector for an image file, the feature vector containing information regarding a likelihood that a selected pair of regions of the image file are of a same intrinsic characteristic, provide a classifier derived from a computer learning technique, compute a classification score for the selected pair of regions of the image file, as a function of the feature vector and the classifier and classify the regions as being of the same intrinsic characteristic, as a function of the classification score, and wherein, when a classification score indicates regions as being of the same intrinsic characteristic, the product includes the further process steps to control a computer to define a constraint between the regions based upon a spatio-spectral relationship defined by the same intrinsic characteristic and to perform an optimization operation as a function of the constraint to utilize the spatio-spectral relationship defined by the same intrinsic characteristic to generate an intrinsic image corresponding to the image, the intrinsic image comprising a representation of one of material or illumination, expressed as a separate, multi-band representation for the one of material or illumination, independent of the other of the material or illumination, wherein each band corresponds to a segment of the electro-magnetic spectrum, such that when the intrinsic image comprises an illumination image, the illumination image captures the intensity and color of light incident upon each point on the surfaces depicted in the image, and when the intrinsic image comprises a material image, the material image captures reflectance properties of surfaces depicted in the image, including the percentage of each wavelength of light a surface reflects, to segregate the image into an intrinsic component.

9. The computer program product of claim 8 wherein the regions comprise adjacent Type C tokens.

10. The computer program product of claim 8 wherein the same intrinsic characteristic comprises a same material.

11. The computer program product of claim 8 wherein the process step to control the computer to provide a classifier derived from a computer learning technique is carried out by providing a process step to control the computer to train the classifier using an image database including coded ground truth versions of images in the image database.

12. The computer program product of claim 11 wherein the coded ground truth versions of the images comprise manually coded images to indicate the same intrinsic characteristic.

13. The computer program product of claim 8 wherein the classifier comprises a classifier arranged to process regions of the image that are adjacent and a classifier arranged to process regions of the image that are non-adjacent.