System and method for detecting objects in images

• US 7,020,337 B2
• Filed: 07/22/2002
• Issued: 03/28/2006
• Est. Priority Date: 07/22/2002
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Jones, Michael J., Viola, Paul A.
• Primary Examiner(s): Mehta, Bhavesh M
• Assistant Examiner(s): STREGE, JOHN B
• Application Number:
  US10/200,464
• Publication Number:
  US 20040013304A1
• Time in Patent Office: 1,345 Days
• Field of Search: 382/115, 382/118, 382/190, 382/195, 382/224, 382/225, 706/12, 706/20
• US Class Current: 382/224
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation

Abstract

A method detects an object, such a face, in an image. The image is first partitioned into patches of various sizes using either an integral image or a Gaussian pyramid. Features in each patch are evaluated to determine a cumulative score. The evaluating is repeated while the cumulative score is within a range of an acceptance threshold and a rejection threshold, and otherwise the image is rejected when the accumulated score is less than the rejection threshold and accepted as including the object when the cumulative score is greater than the acceptance threshold.

Claims

1. A method for detecting an object in an image, comprising:

evaluating a plurality of features in each image to determine a cumulative score, in which the cumulative score is increased by α

_i if ƒ

(ν

_i, I)>

θ

_i, and otherwise increased by β

_i, where ƒ

( ) is a function of a projection ν

_i of a particular patch I_i, θ

_i is a comparison threshold, and α and

β

are acceptance and rejection scores, respectively, and wherein the function is a dot product between ν

_i and I_i expressed as vectors;

repeating the evaluating while the cumulative score is within a range of an acceptance threshold and a rejection threshold; and

otherwise

rejecting the image when the accumulated score is less than the rejection threshold and accepting the image as including the object when the cumulative score is greater than the acceptance threshold.

2. The method of claim 1 further comprising:

partitioning the image into a plurality of patches;

scaling the patches to a plurality of sizes.

3. The method of claim 2 wherein the scaling uses a multi-resolution image.

4. The method of claim 3 wherein the multi-resolution image is a Gaussian pyramid.

5. The method of claim 1 wherein the function ƒ

( ) is non-linear.

6. The method of claim 5 wherein the non-linear function is an absolute value.

7. The method of claim 1 wherein the parameters ν

_i, θ

_i, α

_i and β

_i are determined during a training phase using a first set of labeled images.

8. The method of claim 1 wherein the first set of labeled images includes positive and negative labeled patches.

9. The method of claim 1 wherein the acceptance threshold and the rejection thresholds are determined during a validation phase using a second set of labeled images.

10. The method of claim 1 wherein the object is a face.