Granted Patent B2

US 8,311,116 · App. 12/497,841 · Granted Nov 13, 2012

Abstract

Claims (50)

1. A method comprising:

performing a transform on at least one image to produce a frequency-domain representation of said at least one image;

based on said frequency-domain representation, i) estimating a plurality of motion vectors and ii) identifying a period;

selecting a dominant vector based on i) said estimated motion vectors and ii) said identified period; and

identifying a range for the dominant motion vector based on said identified period,

wherein said selecting the dominant motion vector comprises:

determining a candidate dominant motion vector;

calculating an absolute value of a difference between said candidate dominant motion vector and each of said estimated motion vectors;

comparing each of said absolute values with said range; and

eliminating estimated motion vectors based on said comparing.

2. The method of claim 1 wherein said identifying a period comprises identifying a period from magnitudes of responses at coefficients used to produce said frequency-domain representation.

3. The method of claim 1 wherein said determining a candidate dominant motion vector comprises selecting a candidate dominant motion vector based on an average of dominant motion vectors for neighboring blocks.

4. The method of claim 1 wherein said eliminating occurs when an absolute value falls within said range.

5. The method of claim 1 wherein said identifying a range comprises looking up a range in a table based on said identified period.

6. The method of claim 1 wherein said range is further based on the number of coefficients used to produce said frequency-domain representations.

7. The method of claim 1 further comprising:

filtering each of two images,

wherein said performing a transform comprises performing a Fourier transform on said filtered images to produce a frequency-domain representation of each of said two images.

8. The method of claim 7 wherein said filtering comprises filtering said images through a high-pass filter and a low-pass filter.

9. A method comprising:

performing a transform on at least one image to produce a frequency-domain representation of said at least one image;

based on said frequency-domain representation, i) estimating a plurality of motion vectors and ii) identifying a period; and

selecting a dominant vector based on i) said estimated motion vectors and ii) said identified period;

filtering each of two images, wherein said performing a transform comprises performing a Fourier transform on said filtered images to produce a frequency-domain representation of each of said two images, and wherein said filtering comprises filtering said images through a high-pass filter and a low-pass filter; and

identifying a range for the dominant motion vector based on said identified period,

wherein said selecting the dominant motion vector comprises:

determining a candidate dominant motion vector;

calculating an absolute value of a difference between said candidate dominant motion vector and each of said estimated motion vectors;

comparing each of said absolute values with said range; and

eliminating estimated motion vectors based on said comparing.

10. The method of claim 9 wherein said determining a candidate dominant motion vector comprises selecting a motion vector estimated from a frequency response of said low-pass filter.

11. The method of claim 2 wherein said identifying a range includes identifying a range for the dominant motion vector based on i) said identified period and ii) the number of coefficients in said frequency-domain representation of said image; and

wherein said selecting comprises eliminating one or more of said plurality of estimated motion vectors based on said identified range.

12. A system comprising:

a transform calculator for performing a transform on at least one image to produce a frequency-domain representation of said image;

an estimating calculator for estimating a plurality of motion vectors based on said frequency-domain representation;

a periodic structure detection and elimination module, responsive to said transform calculator and said estimating calculator,

for identifying a period based on said frequency-domain representation of said image, and

for selecting a dominant motion vector based on i) said estimated motion vectors and ii) said identified period; and

a high-pass filter and a low-pass filter, wherein said transform calculator is responsive to said high-pass filter and said low-pass filter;

wherein said identified periodic structure detection and elimination module is configured to:

identify a range for a dominant motion vector in response to said transform calculator;

determine a candidate dominant motion vector;

calculate an absolute value of a difference between said candidate dominant motion vector and each of said estimated motion vectors;

compare each of said absolute values with said range; and

eliminate estimated motion vectors based on said comparing.

13. The system of claim 12 wherein said candidate dominant motion vector is based on an average of dominant motion vectors for neighboring blocks.

14. The system of claim 12 wherein said candidate dominant motion vector is based on a motion vector estimated from a frequency response of said low-pass filter.

15. The system of claim 12 wherein said eliminating occurs when an absolute value falls within said range.

16. The system of claim 12 wherein said identifying a range comprises looking up a range in a table based on said identified period.

Assignments (9)