Probabilistic color classification

Patent metadata

Field	Value
Publication number	US-9563824-B2
Application number	US-201514739635-A
Country	US
Kind code	B2
Filing date	Jun 15, 2015
Priority date	Jun 15, 2015
Publication date	Feb 7, 2017
Grant date	Feb 7, 2017

Abstract

Methods, systems, computer-readable media, and apparatuses for assigning a color class of a defined finite set of colors to at least one sub-region within a test image are presented. A plurality of sub-regions are identified within a test image. A first sub-region color value is determined for a selected first sub-region of the test image. Using the first sub-region color value and a plurality of zero-order probability distributions, a first color class of the defined finite set of colors is determined as a hypothesis color for the first sub-region. A second sub-region color value is determined for a selected second sub-region of the test image. Using the second sub-region color value and a conditional probability distribution conditioned on the hypothesis color for the first sub-region, a second color class of the defined set of colors is determined. The second color class is assigned to the second sub-region.

First claim

What is claimed is:

1. A method for assigning a color class of a defined finite set of colors to at least one sub-region within a test image, comprising, by an image processing device:
receiving the test image from a user device;
identifying a plurality of sub-regions within the test image, wherein pixels within a sub-region of the test image have substantially the same color;
selecting a first sub-region of the plurality of sub-regions, determining a first sub-region color value for the first sub-region;
determining, using the first sub-region color value and a plurality of zero-order probability distributions, a first color class of the defined finite set of colors as a hypothesis color for the first sub-region, wherein the first color class is a highest probability color class for the first sub-region color value;
selecting a second sub-region of the plurality of sub-regions;
determining a second sub-region color value for the second sub-region;
determining, using the second sub-region color value and a conditional probability distribution conditioned on the hypothesis color for the first sub-region, a second color class of the defined finite set of colors to assign to the second sub-region, wherein the second color class is a highest probability color class for the second sub-region color value; and
assigning the second color class to the second sub-region.

2. The method of claim 1 , further comprising:
calculating a first joint probability score associated with the hypothesis color for the first sub-region, using:
a first probability based on:
a zero-order probability distribution associated with the hypothesis color for the first sub-region, and
the first sub-region color value; and

a second probability based on:
the conditional probability distribution conditioned on the hypothesis color for the first sub-region, and
the second sub-region color value.

3. The method of claim 2 , further comprising:
calculating a plurality of joint probability scores, wherein each joint probability score is associated with a hypothesis color determined for a different sub-region of the plurality of sub-regions; and
assigning colors to the plurality of sub-regions based on a highest joint probability score.

4. The method of claim 1 , wherein:
the test image includes an object region corresponding to an object,
the object region includes a plurality of sub-regions corresponding to components of the object, and
each component of the object has a color from the defined finite set of colors.

5. The method of claim 1 , wherein the plurality of zero-order probability distributions are determined using a plurality of training images, wherein a first training image of the plurality of training images is captured in a different lighting environment from a second training image of the plurality of training images.

6. The method of claim 5 , wherein a training image of the plurality of training images includes a training object region corresponding to a training object, wherein a training object includes all colors of the defined finite set of colors.

7. The method of claim 5 , wherein the conditional probability distribution is determined using at least a first zero-order probability distribution and a second zero-order probability distribution of the plurality of zero-order probability distributions.

8. The method of claim 1 , further comprising:
selecting a third sub-region of the plurality of sub-regions;
determining a third sub-region color value for the third sub-region;
determining, using the third sub-region color value and a conditional probability distribution conditioned on:
the hypothesis color for the first sub-region, and
the second color class assigned to the second sub-region,
a third color class of the defined finite set of colors to assign to the third sub-region, wherein the third color class is a highest probability color class for the third sub-region color value; and
assigning the third color class to the third sub-region.

9. A system for assigning a color class of a defined finite set of colors to at least one sub-region within a test image, the system comprising:
a processor; and
a memory coupled to the processor and configurable for storing instructions;
wherein the processor is configured to:
receive the test image from a user device;
identify a plurality of sub-regions within the test image, wherein pixels within a sub-region of the test image have substantially the same color;
select a first sub-region of the plurality of sub-regions,
determine a first sub-region color value for the first sub-region;
determine, using the first sub-region color value and a plurality of zero-order probability distributions, a first color class of the defined finite set of colors as a hypothesis color for the first sub-region, wherein the first color class is a highest probability color class for the first sub-region color value;
select a second sub-region of the plurality of sub-regions;
determine a second sub-region color value for the second sub-region;
determine, using the second sub-region color value and a conditional probability distribution conditioned on the hypothesis color for the first sub-region, a second color class of the defined finite set of colors to assign to the second sub-region, wherein the second color class is a highest probability color class for the second sub-region color value; and
assign the second color class to the second sub-region.

10. The system of claim 9 , wherein the processor is further configured to:
calculate a first joint probability score associated with the hypothesis color for the first sub-region, using:
a first probability based on:
a zero-order probability distribution associated with the hypothesis color for the first sub-region, and
the first sub-region color value; and

a second probability based on:
the conditional probability distribution conditioned on the hypothesis color for the first sub-region, and
the second sub-region color value.

11. The system of claim 10 , wherein the processor is further configured to:
calculate a plurality of joint probability scores, wherein each joint probability score is associated with a hypothesis color determined for a different sub-region of the plurality of sub-regions; and
assign colors to the plurality of sub-regions based on a highest joint probability score.

12. The system of claim 9 , wherein:
the test image includes an object region corresponding to an object,
the object region includes a plurality of sub-regions corresponding to components of the object, and
each component of the object has a color from the defined finite set of colors.

13. The system of claim 9 , wherein, the plurality of zero-order probability distributions are determined using a plurality of training images, wherein a first training image of the plurality of training images is captured in a different lighting environment from a second training image of the plurality of training images.

14. The system of claim 13 , wherein a training image of the plurality of training images includes a training object region corresponding to a training object, wherein a training object includes all colors of the defined finite set of colors.

15. The system of claim 13 , wherein the conditional probability distribution is determined using at least a first zero-order probability distribution and a second zero-order probability distribution of the plurality of zero-order probability distributions.

16. The system of claim 9 , wherein the processor is further configured to:
select a t

Assignees

• Qualcomm Inc

Inventors

• Maggio Emilio
• Pan Qi
• Reitmayr Gerhard

Classifications

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06F18/2415

  based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate · CPC title

• G06V10/56 Primary

  relating to colour · CPC title

• G06F18/24155

  Bayesian classification · CPC title

• G06V10/26

  Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion · CPC title