Iterative fisher linear discriminant analysis
View Patent ↗An exemplary method includes receiving an image data set that comprises a multidimensional property space and data classifiable into data classes, determining a projection vector for data of the data set wherein the projection vector maximizes a ratio of between-class scatter to within-class scatter, selecting a reference for the vector, projecting at least some of the data onto the vector, measuring distances from the reference to at least some of the data, classifying at least some of the data into data classes based on a nesting analysis of the distances, eliminating the classified data from the image data set to produce a modified image data set and deciding whether to determine another projection vector for the modified image data set. Various other exemplary methods, devices, systems, etc., are also disclosed.
1. A method comprising:
receiving an image data set in a multidimensional property space that comprises data classifiable into data classes;
determining a projection vector that maximizes overall variance of the data of the image data set using a processor;
determining a projection vector that maximizes variance between classes of the image data set and minimizes variance within classes of the image data set using the processor;
deciding which vector allows for distinguishing more classes;
projecting at least some of the data onto the decided vector;
classifying at least some of the data into a plurality of unambiguously defined data classes based on the projecting operation, using the processor;
eliminating at least some of the classified unambiguously defined data classes from the image data set based on the classifying operation and using the processor to form a modified image data set containing only unclassified data; and
determining one or more additional projection vectors selected from a group consisting of projection vectors that maximize overall variance and projection vectors that maximize variance between the unambiguous classes and minimize variance within the unambiguous classes.
2. The method of claim 1 wherein the classifying comprises binning distances to form one or more histograms.
3. The method of claim 1 wherein the classifying comprises nesting analysis.
4. The method of claim 1 wherein the classifying comprises binning distances to form one or more histograms and a nesting analysis of the one or more histograms.
5. The method of claim 1 further comprising deciding whether to determine another projection vector that maximizes overall variance of the modified image data set.
6. The method of claim 1 further comprising deciding whether to determine another projection vector that maximizes variance between classes of the modified image data set and minimizes variance within classes of the modified image data set.
7. The method of claim 1 wherein the dimensions of the property space comprises color property dimensions.
8. The method of claim 1 wherein the dimensions of the property space comprise a red property dimension, a green property dimension and a blue property dimension.
9. The method of claim 1 wherein the data classes comprise a data class defined by at least a color vector.
10. One or more computer-readable media storing computer-readable instructions executable by a processor for performing a method comprising:
receiving an image data set in a multidimensional property space that comprises data classifiable into data classes;
determining a projection vector that maximizes overall variance of the data of the image data set using a processor;
determining a projection vector that maximizes variance between classes of the image data set and minimizes variance within classes of the image data set using the processor;
deciding which vector allows for distinguishing more classes;
projecting at least some of the data onto the decided vector;
classifying at least some of the data into a plurality of unambiguously defined data classes based on the projecting operation, using the processor;
eliminating at least some of the classified unambiguously defined data classes from the image data set based on the classifying operation and using the processor to form a modified image data set containing only unclassified data; and
determining one or more additional projection vectors selected from a group consisting of projection vectors that maximize overall variance and projection vectors that maximize variance between the unambiguous classes and minimize variance within the unambiguous classes.
11. The one or more computer-readable media of claim 10 wherein the classifying comprises binning distances to form one or more histograms.
12. The one or more computer-readable media of claim 10 wherein the classifying comprises nesting analysis.
13. The one or more computer-readable media of claim 10 wherein the classifying comprises binning distances to form one or more histograms and a nesting analysis of the one or more histograms.
14. The one or more computer-readable media of claim 10 , the method further comprising deciding whether to determine another projection vector that maximizes overall variance of the modified image data set.
15. The one or more computer-readable media of claim 10 , the method further comprising deciding whether to determine another projection vector that maximizes variance between classes of the modified image data set and minimizes variance within classes of the modified image data set.
16. The one or more computer-readable media of claim 10 wherein the dimensions of the property space comprises color property dimensions.
17. The one or more computer-readable media of claim 10 wherein the dimensions of the property space comprise a red property dimension, a green property dimension and a blue property dimension.
18. The one or more computer-readable media of claim 10 wherein the data classes comprise a data class defined by at least a color vector.