Method for improving classification results of a classifier

• US 10,037,480 B2
• Filed: 03/02/2012
• Issued: 07/31/2018
• Est. Priority Date: 03/04/2011
• Status: Active Grant

• Current Assignee: LBT Innovations Ltd.
• Original Assignee: LBT Innovations Ltd.
• Inventors: van den Hengel, Anton John, Ward, Benjamin William, Hill, Rhys Ernst
• Primary Examiner(s): PARK, CHAN S
• Assistant Examiner(s): Rice, Elisa
• Application Number:
  US14/002,702
• Publication Number:
  US 20140219553A1
• Time in Patent Office: 2,342 Days
• CPC Class Codes: G06F 18/2323 based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
• CPC Class Codes: G06F 18/2431 Multiple classes
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20072 Graph-based image processing
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/162 involving graph-based methods
• CPC Class Codes: G06V 10/7635 based on graphs, e.g. graph cuts or spectral clustering

Abstract

A method for improving classification results of a classifier including receiving classification results for a plurality of elements that have been classified by a classifier as one of a plurality of classes, constructing a graph having a plurality of nodes, each node corresponding to one of the elements, and a plurality of labels, each label corresponding to one of the classes, adding edges between nodes corresponding to related elements, adding edges between each node and each label, and using a graph cut algorithm to cut edges to a node and partition the graph into classes, the graph cut algorithm using as input the classification results for the element corresponding to that node and related elements.

Claims

1. A method for improving classification results of a classifier including:

receiving, using a processor, classification results for a plurality of elements that have been classified by a classifier as one of a plurality of classes;

constructing, using a processor, a graph having a plurality of nodes, each node corresponding to one of the elements, and a plurality of labels, each label corresponding to one of the classes;

adding, using a processor, edges between nodes corresponding to related elements;

adding, using a processor, edges between each node and each label, and

using a graph cut algorithm to cut edges to a node and partition the graph into classes, the graph cut algorithm using as input the classification results for the element corresponding to that node and the related elements for the element corresponding to that node, wherein;

the classifier is a multi-class classifier that has been applied to classify the elements as one of three or more classes;

the classification results include a confidence value for each element, indicating confidence in the classification of that element as each of the three or more classes;

the graph cut algorithm further uses as input the confidence value of the classification for the element corresponding to that node so as to cut edges to a node and partition the graph into the three or more classes; and

,

each element is a pixel in an image of a solid culture medium and any microbial growth on the solid culture medium;

the graph cut algorithm further uses as input the confidence value of the classification for related elements; and

wherein the related elements are adjacent pixels in the image.

2. A method according to claim 1, wherein the classifier is a boosted classifier.

3. A method according to claim 1, wherein the graph cut algorithm further uses as input a likelihood that the related elements have the same classification.

4. A method according to claim 1, wherein the graph cut algorithm includes an alpha expansion procedure.

5. A method according to claim 1, wherein the related elements are eight adjacent pixels.

6. A method according to claim 1, wherein the related elements are four adjacent pixels.

7. A method according to claim 1, wherein each element is a contiguous region in the image.

8. A method according to claim 7, wherein the related elements are adjacent contiguous regions in the image.

9. A method according to claim 7, wherein each contiguous region includes pixels with the same classification and the confidence value is above a threshold.

10. A method according to claim 1, wherein the graph cut algorithm further uses as input an expected size of an object in the image.

11. A method according to claim 1, wherein the graph cut algorithm further uses as input a predetermined rule regarding adjacent pixels.

12. A method according to claim 1, wherein two of the three or more classes include a classification that a pixel in the image is microbial growth and a classification that a pixel in the image is not microbial growth.

13. A method according to claim 12, wherein one class of the three or more classes includes a classification that a pixel in the image is a background corresponding to not microbial growth.

14. A method according to claim 12, wherein one further class of the three or more classes includes a classification that a pixel in the image is a type of the microbial growth.

15. A non-transitory computer readable media including software for use with a computer including a processor and memory for storing the software, the software including a series of instructions executable by the processor to carry out a method for improving classification results of a classifier, the method including:

receiving, using a processor, classification results for a plurality of elements that have been classified by a classifier as one of a plurality of classes,

constructing, using a processor, a graph having a plurality of nodes, each node corresponding to one of the elements, and a plurality of labels, each label corresponding to one of the classes,

adding, using a processor, edges between nodes corresponding to related elements,

adding, using a processor, edges between each node and each label, and

using a graph cut algorithm to cut edges to a node and partition the graph into classes, the graph cut algorithm using as input the classification results for the element corresponding to that node and the related elements for the element corresponding to that node, wherein

the classifier is a multi-class classifier that has been applied to classify the elements as one of three or more classes, and

the classification results include a confidence value for each element, indicating confidence in the classification of that element as each of the three or more classes,

the graph cut algorithm further uses as input the confidence value of the classification for the element corresponding to that node so as to cut edges to a node and partition the graph into the three or more classes, and

wherein each element is a pixel in an image of a solid culture medium and any microbial growth on the solid culture medium;

the graph cut algorithm further uses as input the confidence value of the classification for related elements; and

the related elements are adjacent pixels in the image.

16. An apparatus comprising:

a processor;

a memory; and

software resident in memory accessible to the processor, the software including a series of instructions executable by the processor to carry out a method for improving classification results of a classifier, the method including;

receiving, using a processor, classification results for a plurality of elements that have been classified by a classifier as one of a plurality of classes,

constructing, using a processor, a graph having a plurality of nodes, each node corresponding to one of the elements, and a plurality of labels, each label corresponding to one of the classes,

adding, using a processor, edges between nodes corresponding to related elements,

adding, using a processor, edges between each node and each label, and

using a graph cut algorithm to cut edges to a node and partition the graph into classes, the graph cut algorithm using as input the classification results for the element corresponding to that node and the related elements for the element corresponding to that node, wherein

the classifier is a multi-class classifier that has been applied to classify the elements as one of three or more classes, and

the classification results include a confidence value for each element, indicating confidence in the classification of that element as each of the three or more classes,

the graph cut algorithm further uses as input the confidence value of the classification for the element corresponding to that node so as to cut edges to a node and partition the graph into the three or more classes, and

wherein each element is a pixel in an image of a solid culture medium and any microbial growth on the solid culture medium;

the graph cut algorithm further uses as input the confidence value of the classification for related elements; and

the related elements are adjacent pixels in the image.