Automatic correction of skewing of digital images

Patent metadata

Field	Value
Publication number	US-2017193321-A1
Application number	US-201715408835-A
Country	US
Kind code	A1
Filing date	Jan 18, 2017
Priority date	Oct 8, 2015
Publication date	Jul 6, 2017
Grant date	—

Abstract

A method and system. An initial angle of rotation is determined by sampling test angles of rotation of a captured digital image and analyzing resultant rotated images to determine a resultant rotated image with a highest number of substantially empty lines. The captured digital image is rotated by the initial angle of rotation to generate a first rotated image. A representative line of each line area of multiple text line areas of the first rotated image is generated. A slope of each representative line is calculated. An aggregated slope of the representative lines is determined. The aggregated slope is converted to a refining angle of rotation. The refining angle of rotation is validated by finding points of intersection of lines connecting opposing ends of adjacent representative lines. The first rotated image is rotated by the refining angle of rotation to result in a final rotated image.

First claim

What is claimed is:

1 . A method, said method comprising:
determining an initial angle of rotation by sampling a plurality of test angles of rotation of a captured digital image and analyzing resultant rotated images to determine a resultant rotated image with a highest number of substantially empty lines; rotating the captured digital image by the initial angle of rotation to generate a first rotated image; generating a representative line of each line area of multiple text line areas of the first rotated image; calculating a slope of each representative line; determining an aggregated slope of all of the representative lines in the first rotated image, wherein the aggregated slope is converted to a refining angle of rotation; validating the refining angle of rotation by finding points of intersection of lines connecting opposing ends of adjacent representative lines, and estimating a substantially vertical line through the point of intersection, wherein the difference between the substantially vertical line and a line perpendicular to the aggregated slope indicates the accuracy of the refining angle of rotation; and rotating the first rotated image by the refining angle of rotation to result in a final rotated image.

2 . The method of claim 1 , wherein said determining the initial angle of rotation, includes:
sampling a plurality of test angles at regular angle intervals of rotation.

3 . The method of claim 1 , wherein said determining the initial angle of rotation, includes:
sampling a plurality of test angles in a range of −90 degrees to +90 degrees rotation.

4 . The method of claim 1 , wherein said determining the initial angle of rotation, includes for each resultant rotated image of a sample test angle:
converting the resultant rotated image into a pixel map with each pixel in the pixel map being determined to be empty or non-empty by applying a pixel content threshold; weighting each line of pixels in the pixel map according to a number of non-empty pixels in said each line; determining an empty line in the pixel map as a line of pixels in the pixel map with a weight less than a line content threshold; and counting a number of empty lines in the resultant rotated image.

5 . The method of claim 1 , wherein said determining the initial angle of rotation does not process content of text lines resulting in low processing costs.

6 . The method of claim 1 , said method further comprising:
determining the text line areas of the first rotated image by forming areas of adjacent pixel lines, wherein the adjacent pixel lines are each above a threshold weight of non-empty pixels, and wherein the determined text line areas are separated by at least one pixel line below the threshold weight of non-empty pixels.

7 . The method of claim 1 , wherein said generating the a representative line of each text line area uses a geometrical or statistical method of averaging a text line area.

8 . The method of claim 1 , said method further comprising:
pre-processing the captured digital image to reduce noise by filtering extraneous detail from the captured digital image.

9 . A computer system, comprising a processor, a memory coupled to the processor, and a computer readable storage device coupled to the processor, said storage device containing program code executable by the processor via the memory to implement a method, said method comprising:
determining an initial angle of rotation by sampling a plurality of test angles of rotation of a captured digital image and analyzing resultant rotated images to determine a resultant rotated image with a highest number of substantially empty lines; rotating the captured digital image by the initial angle of rotation to generate a first rotated image; generating a representative line of each line area of multiple text line areas of the first rotated image; calculating a slope of each representative line; determining an aggregated slope of all of the representative lines in the first rotated image, wherein the aggregated slope is converted to a refining angle of rotation; validating the refining angle of rotation by finding points of intersection of lines connecting opposing ends of adjacent representative lines, and estimating a substantially vertical line through the point of intersection, wherein the difference between the substantially vertical line and a line perpendicular to the aggregated slope indicates the accuracy of the refining angle of rotation; and rotating the first rotated image by the refining angle of rotation to result in a final rotated image.

10 . The computer system of claim 9 , wherein said determining the initial angle of rotation, includes:
sampling a plurality of test angles at regular angle intervals of rotation.

11 . The computer system of claim 9 , wherein said determining the initial angle of rotation, includes:
sampling a plurality of test angles in a range of −90 degrees to +90 degrees rotation.

12 . The computer system of claim 9 , wherein said determining the initial angle of rotation, includes for each resultant rotated image of a sample test angle:
converting the resultant rotated image into a pixel map with each pixel in the pixel map being determined to he empty or non-empty by applying a pixel content threshold; weighting each line of pixels in the pixel map according to a number of non-empty pixels in said each line; determining an empty line in the pixel nap as a line of pixels in the pixel map with a weight less than a line content threshold; and counting a number of empty lines in the resultant rotated image.

13 . The computer system of claim 9 , wherein said determining the initial angle of rotation does not process content of text lines resulting in low processing costs.

14 . The computer system of claim 9 , said method further comprising:
determining the text line areas of the first rotated image by forming areas of adjacent pixel lines, wherein the adjacent pixel lines are each above a threshold weight of non-empty pixels, and wherein the determined text line areas are separated by at least one pixel line below the threshold weight of non-empty pixels.

15 . A computer program product, comprising a computer readable storage device having computer readable program code stored therein, said program code containing instructions executable by a processor of a computer system to implement a method, said method comprising:
determining an initial angle of rotation by sampling a plurality of test angles of rotation of a captured digital image and analyzing resultant rotated images to determine a resultant rotated image with a highest number of substantially empty lines; rotating the captured digital image by the initial angle of rotation to generate a first rotated image; generating a representative line of each line area of multiple text line areas of the first rotated image; calculating a slope of each representative line; determining an aggregated slope of all of the representative lines in the first rotated image, wherein the aggregated slope is converted to a refining angle of rotation; validating the refining angle of rotation by finding points of intersection of lines connecting opposing ends of adjacent representative lines, and estimating a substantially vertical line through the point of intersection, wherein the difference between the substantially vertical line and a line perpendicular to the aggregated slope indicates the accuracy of the refining angle of rotation; and rotating the first rotated image by the refining angle of rotation to result in a final rotated i

Assignees

• IBM

Inventors

• Abbas Ahmed

Classifications

• G06V30/1475 Primary

  Inclination or skew detection or correction of characters or of image to be recognised · CPC title

• G06V30/1478

  of characters or characters lines · CPC title

• G06V30/10

  Character recognition · CPC title

• G06K9/00442

  Physics · mapped topic

• G06K9/40

  Physics · mapped topic