1. A method for classifying features in a digital medical image, said method comprising:
using a computer to perform the steps of:
providing a plurality of feature points in an n-dimensional space, wherein said feature points have been extracted from a digital medical image, wherein each feature point is a member of one of two sets;
determining an approximate classifying plane that separates feature points in a first of the two sets from feature points in a second of the two sets;
transforming the classifying plane wherein a normal vector to said transformed classifying plane has positive coefficients and a feature domain for one or more feature points of one of said two sets is a unit hypercube in a transformed space having n axes;
finding a subregion of said unit hypercube containing at least one feature point wherein a transformed rule is extracted;
inversely transforming said transformed rule to obtain a new rule containing one or more feature points of said one of said two sets, wherein said new rule defines a hypercube that is a subset of one of the two feature sets and wherein said hypercube has a vertex that lies in the classifying plane; and
removing the feature points contained by said new rule from said one of two sets.
2. The method of claim 1, further comprising initializing a set of rules to an empty set, and forming a union of said rule set with said new rule.
3. The method of claim 2, wherein if said one of two sets is not empty after removal of said feature points, further comprising, for each of n subregions not covered by said new rule, repeating said steps of transforming the classifying plane, finding a subregion of said unit hypercube, inversely transforming said subregion to obtain a new rule, and removing the feature points contained by said new rule, until said one of two sets is empty, and forming a union of said rule set with each said new rule.
4. The method of claim 3, wherein said steps are repeated for those subregions of said n subregions that contain feature points in said one of two sets.
5. The method of claim 3, wherein each new rule obtained by repeating said steps has zero intersection with each rule in said rule set, and wherein said rule set converges to a complete covering of said feature points in said one of two sets.
6. The method of claim 1, wherein finding a subregion comprises performing a depth first search.
7. The method of claim 1, wherein said transformed rule extracted from said subregion defines a sub-hypercube of said unit hypercube with a largest possible volume.
8. The method of claim 1, wherein said transformed rule extracted from said subregion defines a sub-hypercube of said unit hypercube with a maximum number of feature points in said one of two sets.
9. The method of claim 1, wherein said unit hypercube in said transformed space has a vertex at the origin of the coordinate system of said transformed space.
10. A method for classifying features in an image, said method comprising:
using a computer to perform the steps of:
providing a plurality of feature points in an n-dimensional space, wherein each feature point is a member of one of two sets;
determining an approximate classifying plane that separates feature points in a first of the two sets from feature points in a second of the two sets;
transforming the classifying plane into an n-dimensional transformed space;
finding a subregion of said unit hypercube containing at least one feature point wherein a transformed rule is extracted;
inversely transforming said transformed rule to obtain a new rule containing one or more feature points of said one of said two sets; and
removing the feature points contained by said new rule from said one of two sets;
repeating for each of said n subregions that contain feature points of said first set and are not covered by said new rule, if said first set is not empty after removal of said feature points, said steps of transforming the classifying plane, finding a subregion wherein a transformed rule is extracted, inversely transforming said transformed rule to obtain another new rule, and removing the feature points contained by said new rule, until said first set is empty; and
forming a rule set from a union of each said new rule, wherein each new rule has zero intersection with each other rule in said rule set, and wherein said rule set converges to a complete covering of said feature points in said one of two sets.
11. The method of claim 10, further comprising providing a new feature point, defining a normal from said new feature point to said classifying plane, and defining the intersection of said normal with said classifying plane as a vertex of a hypercube that contains said new feature point at the center of said hypercube.
12. The method of claim 10, wherein said feature points have been extracted from an image.
13. The method of claim 12, wherein said image is a digital medical image.
14. The method of claim 10, wherein said transformed space includes a normal vector to said transformed classifying plane with positive coefficients.
15. The method of claim 10, wherein said transformed space includes a feature domain for one or more feature points of a first of said two sets that is a unit hypercube having n axes with a vertex at the origin.
16. A program storage device readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method steps for classifying features in a digital medical image, said method comprising the steps of:
providing a plurality of feature points in an N-dimensional space, wherein said feature points have been extracted from a digital medical image, wherein each feature point is a member of one of two sets;
determining an approximate classifying plane that separates feature points in a first of the two sets from feature points in a second of the two sets;
transforming the classifying plane wherein a normal vector to said transformed classifying plane has positive coefficients and a feature domain for one or more feature points of one of said two sets is a unit hypercube in a transformed space having n axes;
finding a subregion of said unit hypercube containing at least one feature point wherein a transformed rule is extracted;
inversely transforming said transformed rule to obtain a new rule containing one or more feature points of said one of said two sets, wherein said new rule defines a hypercube that is a subset of one of the two feature sets and wherein said hypercube has a vertex that lies in the classifying plane; and
removing the feature points contained by said new rule from said one of two sets.
17. The program storage device readable by a computer of claim 16, the method further comprising initializing a set of rules to an empty set, and forming a union of said rule set with said new rule.
18. The program storage device readable by a computer of claim 17, wherein if said one of two sets is not empty after removal of said feature points, further comprising, for each of n subregions not covered by said new rule, repeating said steps of transforming the classifying plane, finding a subregion of said unit hypercube, inversely transforming said subregion to obtain a new rule, and removing the feature points contained by said new rule, until said one of two sets is empty, and forming a union of said rule set with each said new rule.
19. The program storage device readable by a computer of claim 18, wherein said steps are repeated for those subregions of said n subregions that contain feature points in said one of two sets.
20. The program storage device readable by a computer of claim 18, wherein each new rule obtained by repeating said steps has zero intersection with each rule in said rule set, and wherein said rule set converges to a complete covering of said feature points in said one of two sets.
21. The program storage device readable by a computer of claim 16, wherein finding a subregion comprises performing a depth first search.
22. The program storage device readable by a computer of claim 16, wherein said transformed rule extracted from said subregion defines a sub-hypercube of said unit hypercube with a largest possible volume.
23. The program storage device readable by a computer of claim 16, wherein said transformed rule extracted from said subregion defines a sub-hypercube of said unit hypercube with a maximum number of feature points in said one of two sets.
24. The program storage device readable by a computer of claim 16, wherein said unit hypercube in said transformed space has a vertex at the origin of the coordinate system of said transformed space.
25. The method of claim 6, further comprising using a predetermined depth parameter to limit a depth of said search.
26. The program storage device readable by a computer of claim 21, wherein said method further comprises using a predetermined depth parameter to limit a depth of said search.
27. The method of claim 10, wherein each said rule defines a hypercube that is a subset of the one of two feature sets and wherein each said hypercube has a vertex that lies in the classifying plane.
The claims below are in addition to those above.
All refrences to claim(s) which appear below refer to the numbering after this setence.
1. A method for fabricating the light emitting diode element comprising steps:
(a) providing a substrate, forming a passivation layer on said substrate and defining a plurality of polygonal etch areas;
(b) etching said substrate to form on said etch areas a plurality of basins, each basin including inclined natural crystal planes inwardly extended from a surface of said substrate and a bottom plane, parallel to said surface of said substrate, connected to the inclined planes, and pattern-etching said bottom plane of said basin to obtain a rugged surface;
(c) forming a light emitting diode structure on said bottom plane of said basin via epitaxially growing on said bottom plane in said basin an n-type III-V group compound layer, an active layer and a p-type III-V group compound layer in sequence, wherein said active layer is interposed between said n-type III-V group compound layer and said p-type III-V group compound layer and functions as a light emitting zone;
(d) vapor-depositing a p-type ohmic contact metal layer on said p-type III-V group compound layer, and connecting said p-type ohmic contact metal layer with a heat-conduction substrate;
(e) removing said substrate;
(f) vapor-depositing an n-type ohmic contact metal layer on said n-type III-V group compound layer to allow the said n-type III-V group compound layer to be interposed between said n-type ohmic contact electrode and said substrate; and
(g) cutting and splitting the total structure into light emitting diode chips.
2. The method for fabricating the light emitting diode element according to claim 1, wherein said substrate is made of sapphire, silicon carbide (SiC), silicon (Si), gallium arsenide (GaAs) or aluminum nitride (AlN).
3. The method for fabricating the light emitting diode element according to claim 1, wherein said passivation layer has a width of between 5 and 50 microns, and said etch area has an inner diameter of between 200 and 2000 microns.
4. The method for fabricating the light emitting diode element according to claim 1, wherein said etch area has a shape of a rectangle, circle, triangle, star, or polygon.
5. The method for fabricating the light emitting diode element according to claim 1, wherein said basin has a depth of between 0.5 and 50 microns.
6. The method for fabricating the light emitting diode element according to claim 1, wherein the material of said heat-conduction substrate is selected from the group consisting of gold (Au), aluminum (Al), copper (Cu), silicon (Si), gallium phosphide (GaP), silicon carbide (SiC); alternatively, the material of said heat-conduction substrate is one combination of the abovementioned materials.
7. The method for fabricating the light emitting diode element according to claim 1, wherein said substrate is removed with a wet-etching method, a dry-etching method, a laser lift off method, or a method of using different thermal expansion coefficients, wherein said substrate is spontaneously separated from said light emitting diode structure during temperature variation.