Resource-Aware Computer Vision

• US 20110305392A1
• Filed: 06/09/2010
• Published: 12/15/2011
• Est. Priority Date: 06/09/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Kapoor, Ashish, Vijayanarasimhan, Sudheendra N.
• Granted Patent: US 8,811,743 B2
• Time in Patent Office: Days
• US Class Current: 382/174
• CPC Class Codes: G06F 18/211 Selection of the most significant subset of features
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
• CPC Class Codes: G06V 10/771 Feature selection, e.g. selecting representative features from a multi-dimensional feature space

Abstract

The described implementations relate to computer vision. In one case image data is received. Resource constraints associated with processing the image data are evaluated. Further, a visual recognition and detection processing strategy is selected for the image data based at least in part on the evaluated resource constraints.

Claims

1. A system, comprising:

a feature extraction component configured to receive an image and to identify parts of the image for further evaluation;

a pattern recognition component configured to recognize object portions from the image parts; and

,

a parameter-based decision component configured to influence the feature extraction component and the pattern recognition component based upon at least one computational resource constraint.

2. The system of claim 1, wherein the pattern recognition component is further configured to identify a feature that is useful for recognizing an individual object portion.

3. The system of claim 1, wherein the parameter-based decision component is further configured to evaluate at least one computational resource prior to image processing by the feature extraction component and the pattern recognition component and to allocate the computational resource to the feature extraction component and the pattern recognition component based upon the evaluation.

4. The system of claim 1, wherein the feature extraction component is configured to map the image utilizing a grid model to identify the image parts as grid parts.

5. The system of claim 2, wherein the feature extraction component is configured to access training data that identifies a feature that is indicative of an individual object portion contained in an individual grid part.

6. The system of claim 1, wherein the at least one computational resource constraint relates to output from one or more of:

a time constraint parameter, a processing constraint parameter, or a memory parameter.

7. The system of claim 1, wherein the parameter-based decision component is configured to influence the feature extraction component relative to determining individual image parts and a number of iterations to examine the individual image parts.

8. The system of claim 1, wherein the parameter-based decision component is configured to utilize a value-of-information obtained per processing cycle by the feature extraction component and the pattern recognition component as the at least one parameter.

9. The system of claim 1, further comprising a set of training images that is accessible by the feature extraction component and the pattern recognition component and that relates individual features that are indicative of individual object portions.

10. A computer-readable storage media having instructions stored thereon that when executed by a computing device cause the computing device to perform acts, comprising:

receiving image data;

evaluating computational resource constraints associated with processing the image data; and

,

selecting a visual recognition and detection processing strategy for the image data based at least in part on the evaluated computational resource constraints.

11. The computer-readable storage media of claim 10, wherein the evaluating and selecting are dynamically repeated in an iterative manner until the processing is complete.

12. The computer-readable storage media of claim 10, wherein the evaluating is performed at build time prior to the receiving, the evaluating is performed in real time after the receiving, or the evaluating is performed prior to the receiving and is re-performed after the receiving.

13. The computer-readable storage media of claim 10, wherein selecting a visual recognition and detection processing strategy comprises selecting feature locations for the image data and feature types for individual feature locations.

14. A computer-readable storage media having instructions stored thereon that when executed by a computing device cause the computing device to perform acts, comprising:

evaluating an image to determine an individual image part to examine for the presence of a portion of an object; and

,

selecting a feature type that is indicative of the object portion, wherein the evaluating and selecting are performed in consideration of available computational resources.

15. The computer-readable storage media of claim 14, further comprising iteratively repeating the evaluating and selecting and comparing a resultant value-of-information obtained to at least one computational resource constraint.

16. The computer-readable storage media of claim 14, wherein the evaluating comprises associating a weight parameter with each individual image part, wherein the weight parameter reflects an importance of an individual object portion for detection of the object.

17. The computer-readable storage media of claim 14, further comprising modeling an appearance of an individual object portion based upon evidence from local features collected within an individual grid part.

18. The computer-readable storage media of claim 14, wherein the selecting comprises examining the individual portions for a feature.

19. The computer-readable storage media of claim 14, wherein the selecting comprises mapping the image into grid parts and wherein individual grid parts appear at a particular location with respect to the object'"'"'s extent, and wherein a weight of an individual grid part provides a measure of importance of analyzing the individual grid part for an individual object portion.

20. The computer-readable storage media of claim 14, wherein the evaluating and the selecting comprise accessing training data that indicates object portions that are relatively highly indicative of the object'"'"'s presence.