Method for Image Skew Detection

• US 20100067826A1
• Filed: 09/15/2009
• Published: 03/18/2010
• Est. Priority Date: 09/18/2008
• Status: Active Grant

• Current Assignee: Vakxel Corporation
• Original Assignee: Certifi Media, Inc.
• Inventors: Honsinger, Chris W., Jones, Paul W.
• Granted Patent: US 8,160,393 B2
• Time in Patent Office: Days
• US Class Current: 382/280
• CPC Class Codes: G06V 10/243 by compensating for image skew or non-uniform image deformations

Abstract

A method for calculating a skew angle of an original image, executed at least in part on a computer system stores image data for the original image in an electronic memory, then forms an energy-normalized image according to the relative contrast amplitude of image features over each of a plurality of local image regions within the stored image data. A partitioned image is formed by partitioning the energy-normalized image into a number of sub-regions. A summed region is formed as a combination of image pixel data from the sub-regions. A Fourier magnitude spectrum is obtained by performing a Fourier transform on the summed region. The skew angle is calculated according to the peak value of a radial line integration function that is formed by integrating the Fourier magnitude spectrum along each of a plurality of lines of constant radial angles. An output signal indicates the calculated skew angle.

Claims

1. A method for calculating a skew angle of an original image, the method executed at least in part on a computer system and comprising:

storing image data for the original image in an electronic memory;

forming an energy-normalized image according to the relative contrast amplitude of image features over each of a plurality of local image regions within the stored image data;

forming a partitioned image by partitioning the energy-normalized image into a plurality of sub-regions;

forming a summed region as a combination of image pixel data from the plurality of sub-regions;

obtaining a Fourier magnitude spectrum by performing a Fourier transform on the summed region;

calculating the skew angle according to a peak value of a radial line integration function that is formed by integrating the Fourier magnitude spectrum along each of a plurality of lines of constant radial angles; and

providing an output signal that is indicative of the calculated skew angle.

2. The method of claim 1 further comprising rotating the original image according to the calculated skew angle.

3. The method of claim 1 further comprising preprocessing the stored image data to reduce image resolution.

4. The method of claim 1 further comprising preprocessing the stored image data by averaging pixel values over one or more regions of the image.

5. The method of claim 1 wherein providing the output signal comprises providing an encoded digital value.

6. The method of claim 1 wherein forming the summed region comprises adding corresponding pixel data values from each of the plurality of sub-regions.

7. The method of claim 1 further comprising providing a confidence-level signal that is indicative of a confidence level related to the calculated skew angle.

8. The method of claim 1 wherein forming the energy-normalized image comprises normalizing the mean and variance over the local image regions.

9. The method of claim 1 wherein forming the energy-normalized image comprises normalizing the minimum-to-maximum pixel contrast over each of the plurality of image regions.

10. The method of claim 1 wherein forming the partitioned image comprises forming the plurality of sub-regions of pixels, the sub-regions of size P×

P, wherein P is an integer.

11. The method of claim 10 further comprising multiplying pixel data values for one or more sub-regions by a scalar.

12. The method of claim 10 wherein P is a power of 2.