Content Recognizer via Probabilistic Mirror Distribution

• US 20060110006A1
• Filed: 12/19/2005
• Published: 05/25/2006
• Est. Priority Date: 06/28/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Venkatesan, Ramarathnam, Johnson, Mark C.
• Granted Patent: US 7,136,535 B2
• Time in Patent Office: Days
• US Class Current: 382/100
• CPC Class Codes: G06V 10/507 Summing image-intensity values; Histogram projection analysis

Abstract

An implementation of a technology, described herein, for facilitating the recognition of content of digital goods. At least one implementation, described herein, derives a probabilistic mirror distribution of a digital good (e.g., digital image or audio signal). It uses the resulting data to derive weighting factors (i.e., coefficients) for the digital good. Based, at least in part on such weighting factors, it determines statistics of the good and quantizes it. The scope of the present invention is pointed out in the appending claims.

Claims

1. A method comprising:

obtaining a digital good;

segmenting the good into a plurality of regions; and

producing a recognition indication for the digital good based upon a combination of quantizations q_R for each region of the plurality.

2. The method of claim 1, further comprising, prior to said producing a recognition indication, generating weighting factor mij for one or more points xij of each region of the plurality of regions.

3. The method of claim 2, wherein the generating comprises:

defining one or more sub-regions Sij inside of a region R;

computing suitable statistics for sub-region Sij;

quantizing the statistics for region R; and

generating weighting factor mij for one or more points xij of region R.

4. The method of claim 2, further comprising producing a quantization qR of each region of the plurality of regions based upon one or more weighting factors mij of one or more points xij.

5. The method of claim 4, wherein the producing a quantization qR comprises hashing an unknown region M using xij as weights, where region M is based upon the weighting factor mij.

6. The method of claim 1, wherein the segmenting generates multiple, pseudorandomly sized and distributed regions of the plurality.

7. The method of claim 1, wherein one or more of the regions overlap.

8. A method facilitating the recognition of content of digital goods, the method comprising:

obtaining a digital good;

segmenting the good into a plurality of pseudorandomly sized and pseudorandomly distributed regions; and

producing a recognition indication for the digital good.

9. The method of claim 8, wherein one or more of the regions overlap.

10. The method of claim 8, wherein the regions are pseudorandomly sized and pseudorandomly distributed according to a specified distribution.

11. The method of claim 8 wherein the segmenting further comprises performing transformations on the good or one or more regions of the good to place the good or one or more regions of the good into a canonical form.

12. The method of claim 11, wherein performing transformations comprises employing discrete wavelet transformation (DWT).

13. The method of claim 8, further comprising, prior to said producing a recognition indication, generating weighting factor mij for one or more points xij of each region of the plurality of distributed regions.

14. The method of claim 13, wherein the generating comprises:

defining one or more sub-regions Sij inside of a region R;

computing suitable statistics for sub-region Sij;

quantizing the statistics for region R; and

generating weighting factor mij for one or more points xij of region R.

15. The method of claim 13, further comprising producing a quantization qR of each region of the plurality of regions based upon one or more weighting factors mij of one or more points xij.

16. A computer-readable medium having computer-executable instructions that, when executed by a computer, performs a method facilitating the recognition of content of digital goods, the method comprising:

for each region of a discreet good that has been segmented into a plurality of regions;

defining one or more sub-regions Sij inside of a region R;

computing suitable statistics for sub-region Sij;

quantizing the statistics for region R;

generating weighting factor mij for one or more points xij of region R; and

producing a quantization qR of region R based upon one or more weighting factors mij of one or more points xij;

producing a recognition indication based upon a combination of the quantizations.

17. The medium recited of claim 16, wherein a sub-region Sij is centered on xij.

18. The medium recited of claim 16, wherein xij is the center of region R.

19. The medium recited of claim 16, wherein a suitable statistic is a finite order moment of a region or sub-region.

20. The medium recited of claim 16, wherein each region R has been segmented such that it is pseudorandomly sized and pseudorandomly distributed.