Automatic correction of digital image distortion

• US 8,285,077 B2
• Filed: 07/15/2008
• Issued: 10/09/2012
• Est. Priority Date: 07/15/2008
• Status: Active Grant

• Current Assignee: Kofax Incorporated
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Feró, László, Solymosi, Pál, Tamási, György
• Primary Examiner(s): ENTEZARI, MICHELLE M
• Application Number:
  US12/173,501
• Publication Number:
  US 20100014782A1
• Time in Patent Office: 1,547 Days
• Field of Search: 382289-290
• US Class Current: 382/290
• CPC Class Codes: G06V 10/243 by compensating for image skew or non-uniform image deformations
• CPC Class Codes: H04N 1/387 Composing, repositioning or otherwise geometrically modifying originals

Abstract

A method of de-skewing a digital image is described. An input camera image is initially received, and text within the input camera image is identified. A text direction of the identified text is determined to determine text lines within the camera image. A three-dimensional de-skewing transformation is determined of the text lines to make the text lines horizontal. Then the de-skewing transformation is applied to the input camera image to form a de-skewed output image. An unwarping transformation may also be applied to the input camera image for straightening text lines that are curved.

Claims

1. A method of de-skewing a digital image affected by perspective distortion, the method comprising:

receiving a skewed input camera image affected by perspective distortion and containing curved lines of text;

identifying text within the input camera image;

determining a text direction of the lines of text to determine text direction lines including establishing a horizontal text direction based on calculation of a horizontal vanishing point and establishing a vertical text direction based on calculation of a vertical vanishing point;

applying a three-dimensional de-skewing transformation of the text direction lines without reference to information from any other image; and

applying a three-dimensional de-skewing transformation of the text direction lines without reference to information from any other image to straighten the curved lines of text and produce a de-skewed output image containing straight horizontal lines of text without perspective distortion.

2. A method according to claim 1, wherein identifying text includes determining black connected components within the input camera image.

3. A method according to claim 2, wherein identifying text includes identifying and ignoring non-text components in the input camera image.

4. A computer program product in a non-transitory computer readable storage medium for de-skewing a digital image affected by perspective distortion, the product comprising:

program code for receiving a skewed input camera image affected by perspective distortion and containing curved lines of text;

program code for identifying text within the input camera image;

program code for determining a text direction of the lines of text to determine text direction lines within the camera image including establishing a horizontal text direction based on calculation of a horizontal vanishing point and establishing a vertical text direction based on calculation of a vertical vanishing point;

program code for applying a three-dimensional de-skewing transformation of the text direction lines without reference to information from any other image; and

program code for applying a three-dimensional de-skewing transformation of the text direction lines without reference to information from any other image to straighten the curved lines of text and produce a de-skewed output image containing straight horizontal lines of text without perspective distortion.

5. A product according to claim 4, wherein the program code for identifying text includes program code for determining black connected components within the input camera image.

6. A product according to claim 5, wherein the program code for identifying text includes program code for identifying and ignoring non-text components in the input camera image.