EXTRACTING DOMINANT COLORS FROM IMAGES USING CLASSIFICATION TECHNIQUES

• US 20080075360A1
• Filed: 09/21/2006
• Published: 03/27/2008
• Est. Priority Date: 09/21/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ma, Wei-Ying, Li, Mingjing, Li, Zhiwei, Chen, Yuanhao
• Granted Patent: US 7,809,185 B2
• Time in Patent Office: Days
• US Class Current: 382/155
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 10/56 relating to colour

Abstract

A method and system for generating a detector to detect a dominant color of an image is provided. A dominant color system trains a detector to classify colors as being dominant colors of images. The dominant color system trains the detector using a collection of training images. To train the detector, the dominant color system first identifies candidate dominant colors of the training images. The dominant color system then extracts features of the candidate dominant colors. The dominant color system also inputs an indication of dominance of each of the candidate dominant colors. The dominant color system then trains a detector to detect the dominant color of images using the extracted features and indications of dominance of the candidate dominant colors as training data.

Claims

1. A method in a computing system for generating a detector to detect a dominant color of an image, the method comprising:

providing a collection of training images;

for each training image,

identifying candidate dominant colors of the training image;

extracting features of the candidate dominant colors of the training image; and

inputting indications of dominance of the candidate dominant colors to the training image; and

training a detector to detect a dominant color of the image using the extracted features and indications of dominance of the candidate dominant colors.

2. The method of claim 1 including:

identifying candidate dominant colors of a target image;

extracting features of the candidate dominant colors of the target image; and

detecting a dominant color of the target image by applying the detector to the extracted features of the target image.

3. The method of claim 1 wherein the identifying of candidate dominant colors of the training image includes

quantizing the colors of the training image into bins;

calculating average colors of the bins;

calculating weights of the bins; and

clustering the average colors of the bins factoring in the weight of the bins, the clustered average colors representing the candidate dominant colors.

4. The method of claim 3 including filtering out clustered average colors based on their weights.

5. The method of claim 3 wherein the clustering includes successively creating a combined color for a pair of colors whose difference is smallest.

6. The method of claim 5 wherein the combined color is a weighted combination of the pair of colors and has a combined weight derived from the weights of the pair of colors.

7. The method of claim 1 wherein the extracting of features of the candidate dominant colors of the training image includes extracting a feature based on difference between a candidate dominant color and colors of the training image.

8. The method of claim 7 wherein the difference is based on an area of the training image selected from a group consisting of an entire area, a central area, and an attention area.

9. The method of claim 1 wherein the extracting of features of the candidate dominant colors of the training image includes extracting a feature based on difference between a similar color of the training image that is similar to a candidate dominant color and colors of neighboring pixels of the pixel with the similar color.

10. The method of claim 1

wherein the identifying of candidate dominant colors of the training image includes:

quantizing the colors of the training image into bins;

calculating average colors of the bins;

calculating weights of the bins; and

clustering the average colors of the bins factoring in the weight of the bins, the clustered average colors representing the candidate dominant colors; and

wherein the extracting of features of the candidate dominant colors of the training image includes extracting a feature based on difference between a candidate dominant color and colors of the training image; and

including;

identifying candidate dominant colors of a target image;

extracting features of the target image; and

detecting a dominant color of the target image by applying the detector to the extracted features of the target image.

11. A computing system that generates a detector to detect a dominant color of an image, comprising:

an identify candidate dominant colors component that identifies candidate dominant colors of an image;

an extract features of candidate dominant colors component that extracts features of candidate dominant colors of an image;

an input indications of dominance component that inputs indications of dominance of identified candidate colors; and

a training component that trains a detector to detect a dominant color of an image using extracted features for candidate dominant colors of training images and indications of dominance of the candidate dominant colors of the training images.

12. The computing system of claim 11 wherein the training component uses the identify candidate dominant colors component to identify candidate dominant colors of the training images, the extract features of candidate dominant colors component to extract features of the candidate dominant colors of the training images, and the input indications of dominance component to input indications of the dominance of the candidate dominant colors of the training images.

13. The computing system of claim 11 including:

an identify dominant color component that identifies a dominant color of a target image by applying the detector to the extracted features of the target image.

14. The computing system of claim 11 including:

an identify dominant color component that identifies a dominant color of a target image by using

the identify candidate dominant colors component to identify candidate dominant colors of the target image;

the extract features of candidate dominant colors component to extract features of candidate dominant colors of the target image; and

the detector to detect a dominant color of the target image using the extracted features of the candidate dominant colors of the target image.

15. The computing system of claim 11 wherein the identify candidate dominant colors component

quantizes the colors of the training image into bins;

calculates average colors of the bins; and

clusters the average colors of the bins, the clustered average colors representing the candidate dominant colors.

16. The computing system of claim 11 wherein the extract features of candidate dominant colors component extracts a feature based on difference between a candidate dominant color and colors of the training image.

17. The computing system of claim 11 wherein the difference is based on an area of the training image selected from a group consisting of an entire area, a central area, and an attention area.

18. A computer-readable medium containing instructions for controlling a computing system to identify a dominant color of a target image, by a method comprising:

identifying candidate dominant colors of the target image;

extracting features of the candidate dominant colors of the target image; and

detecting a dominant color of the target image by applying a detector to the extracted features of the candidate dominant colors, the detector to be generated by

identifying candidate dominant colors of training images;

extracting features of the candidate dominant colors of the training images;

receiving indications of dominance of the candidate dominant colors of the training images; and

training the detector to detect a dominant color of an image using the extracted features and indications of dominance of the candidate dominant colors of the training images.

19. The computer-readable medium of claim 18 wherein the identifying of candidate dominant colors includes clustering similar colors of an image.

20. The computer-readable medium of claim 19 wherein the extracting of features of the candidate dominant colors includes extracting a feature based on difference between a candidate dominant color and colors of an image.