Detecting an object within an image by incrementally evaluating subwindows of the image in parallel

• US 8,077,920 B2
• Filed: 05/19/2010
• Issued: 12/13/2011
• Est. Priority Date: 11/24/2004
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Bourdev, Lubomir D.
• Primary Examiner(s): Dang, Duy M
• Application Number:
  US12/783,286
• Publication Number:
  US 20100226583A1
• Time in Patent Office: 573 Days
• Field of Search: 382/103, 382/150, 382/173, 382/176, 382/190, 382/205, 382/225, 348/169
• US Class Current: 382/103
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/776 Validation; Performance evaluation

Abstract

One embodiment of the present invention provides a system for detecting an occurrence of an object in an image. During operation, the system selects a subwindow to be evaluated based upon a currently estimated likelihood of the object being detected within the subwindow. The system then performs an evaluation step on the subwindow to determine if the object exists in the subwindow, wherein performing the evaluation step involves updating the currently estimated likelihood of the object being detected in the subwindow. If evaluation of the subwindow is not complete after performing the evaluation step, the system stores information associated with the subwindow to facilitate subsequent evaluation of the subwindow.

Claims

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

performing, by a computer;

generating a plurality of scan windows, wherein each of the plurality of scan windows comprises a respective subwindow of the image;

performing an initial evaluation step for each of the plurality of scan windows, wherein performing the initial evaluation step for a respective one of the plurality of scan windows comprises determining and storing an estimated likelihood of the object being detected in the respective one of the plurality of scan windows, and wherein a first scan window of the plurality of scan windows has a highest estimated likelihood of the object being detected;

selecting the first scan window for a further evaluation step based on the first scan window having the highest estimated likelihood of the object being detected; and

performing the further evaluation step for the first scan window, wherein performing the further evaluation step for the first scan window comprises determining and storing an updated estimated likelihood of the object being detected in the first scan window.

2. The method of claim 1, further comprising:

selecting the first scan window for a full evaluation step based on the updated estimated likelihood of the object being detected in first scan window being higher than a first threshold value; and

performing the full evaluation step for the first scan window, wherein performing the full evaluation step for the first scan window comprises determining whether the object is detected in the first scan window.

3. The method of claim 1, wherein the initial evaluation step is performed in parallel for at least two of the plurality of scan windows.

4. The method of claim 1, further comprising:

storing information indicative of each of the plurality of scan windows and their respective estimated likelihoods in a data structure; and

sorting the data structure by the respective estimated likelihoods of the plurality of scan windows.

5. The method of claim 4, further comprising:

discarding the first scan window from the data structure based on the updated estimated likelihood of the object being detected in the first scan window being lower than a second threshold value.

6. The method of claim 4, further comprising:

returning the first scan window to the data structure based on the updated estimated likelihood of the object being detected in the first scan window being lower than a first threshold value and higher than a second threshold value.

7. The method of claim 6, further comprising:

selecting a second scan window of the plurality of scan windows for the further evaluation step based on the estimated likelihood of the object being detected in the second scan window; and

performing the further evaluation step for the second scan window, wherein performing the further evaluation step for the second scan window comprises determining and storing an updated estimated likelihood of the object being detected in the second scan window.

8. A non-transitory computer-readable storage device storing instructions that when executed by a computer cause the computer to perform a method for detecting an object in an image, the method comprising:

generating a plurality of scan windows, wherein each of the plurality of scan windows comprises a respective subwindow of the image;

performing an initial evaluation step for each of the plurality of scan windows, wherein performing the initial evaluation step for a respective one of the plurality of scan windows comprises determining and storing an estimated likelihood of the object being detected in the respective one of the plurality of scan windows, and wherein a first scan window of the plurality of scan windows has a highest estimated likelihood of the object being detected;

selecting the first scan window for a further evaluation step based on the first scan window having the highest estimated likelihood of the object being detected; and

performing the further evaluation step for the first scan window, wherein performing the further evaluation step for the first scan window comprises determining and storing an updated estimated likelihood of the object being detected in the first scan window.

9. The non-transitory computer-readable storage device of claim 8, wherein the method further comprises:

selecting the first scan window for a full evaluation step based on the updated estimated likelihood of the object being detected in first scan window being higher than a first threshold value; and

performing the full evaluation step for the first scan window, wherein performing the full evaluation step for the first scan window comprises determining whether the object is detected in the first scan window.

10. The non-transitory computer-readable storage device of claim 8, wherein the initial evaluation step is performed in parallel for at least two of the plurality of scan windows.

11. The non-transitory computer-readable storage device of claim 8, wherein the method further comprises:

storing information indicative of each of the plurality of scan windows and their respective estimated likelihoods in a data structure; and

sorting the data structure by the respective estimated likelihoods of the plurality of scan windows.

12. The non-transitory computer-readable storage device of claim 11, wherein the method further comprises:

discarding the first scan window from the data structure based on the updated estimated likelihood of the object being detected in the first scan window being lower than a second threshold value.

13. The non-transitory computer-readable storage device of claim 11, wherein the method further comprises:

returning the first scan window to the data structure based on the updated estimated likelihood of the object being detected in the first scan window being lower than a first threshold value and higher than a second threshold value.

14. The non-transitory computer-readable storage device of claim 13, wherein the method further comprises:

selecting a second scan window of the plurality of scan windows for the further evaluation step based on the estimated likelihood of the object being detected in the second scan window; and

performing the further evaluation step for the second scan window, wherein performing the further evaluation step for the second scan window comprises determining and storing an updated estimated likelihood of the object being detected in the second scan window.

15. An apparatus that detects an object in an image, comprising:

one or more processors; and

a memory coupled to the one or more processors and storing program instructions executable by the one or more processors to implement;

generating a plurality of scan windows, wherein each of the plurality of scan windows comprises a respective subwindow of the image;

performing an initial evaluation step for each of the plurality of scan windows, wherein performing the initial evaluation step for a respective one of the plurality of scan windows comprises determining and storing an estimated likelihood of the object being detected in the respective one of the plurality of scan windows, and wherein a first scan window of the plurality of scan windows has a highest estimated likelihood of the object being detected;

selecting the first scan window for a further evaluation step based on the first scan window having the highest estimated likelihood of the object being detected; and

performing the further evaluation step for the first scan window, wherein performing the further evaluation step for the first scan window comprises determining and storing an updated estimated likelihood of the object being detected in the first scan window.

16. The apparatus of claim 15, wherein the program instructions are executable by the one or more processors to implement:

selecting the first scan window for a full evaluation step based on the updated estimated likelihood of the object being detected in first scan window being higher than a first threshold value; and

performing the full evaluation step for the first scan window, wherein performing the full evaluation step for the first scan window comprises determining whether the object is detected in the first scan window.

17. The apparatus of claim 15, wherein the initial evaluation step is performed in parallel for at least two of the plurality of scan windows.

18. The apparatus of claim 15, wherein the program instructions are executable by the one or more processors to implement:

storing information indicative of each of the plurality of scan windows and their respective estimated likelihoods in a data structure; and

sorting the data structure by the respective estimated likelihoods of the plurality of scan windows.

19. The apparatus of claim 18, wherein the program instructions are executable by the one or more processors to implement:

discarding the first scan window from the data structure based on the updated estimated likelihood of the object being detected in the first scan window being lower than a second threshold value.

20. The apparatus of claim 18, wherein the program instructions are executable by the one or more processors to implement:

returning the first scan window to the data structure based on the updated estimated likelihood of the object being detected in the first scan window being lower than a first threshold value and higher than a second threshold value.