Method and device for extracting distorted straight line from image

Patent metadata

Field	Value
Publication number	US-9317766-B2
Application number	US-201414563335-A
Country	US
Kind code	B2
Filing date	Dec 8, 2014
Priority date	Dec 9, 2013
Publication date	Apr 19, 2016
Grant date	Apr 19, 2016

Abstract

A method and a device for extracting a distorted straight line from an image. The method includes performing a line segment detection process on the image; for each line segment detected, finding adjacent line segments having a low non-collinear degree with a current line segment, so as to combine them together to obtain one or more line segment combinations; and extracting the distorted straight line according to the line segment combination.

First claim

What is claimed is:

1. A method for extracting a distorted straight line from an image, comprising:
performing a line segment detection process on the image;
for each line segment detected, finding adjacent line segments having a low non-collinear degree with a current line segment, so as to combine them together to obtain one or more line segment combinations; and
extracting the distorted straight line according to the line segment combination,
wherein finding adjacent line segments having a low non-collinear degree with a current line segment further comprises:
defining an adjoining rectangular area along a slope direction of the current line segment; and
searching another line segment whose non-collinear degree with a current line segment is lower than a predetermined threshold as the adjacent line segment within the adjoining rectangular area in accordance with a principle:
the middle first, then both sides; and
from near to far,

wherein the non-collinear degree is measured by calculating a connection angle with the following formula, the less the connection angle, the lower the non-collinear degree:
connection angle=max(abs( a 2− a 1),abs( a 2− a 3)); wherein

a 1 is an angle between the current line segment and a predetermined reference line;
a 2 is an angle between a connection line of the current line segment with the another line segment and the predetermined reference line; and
a 3 is an angle between the another line segment and the predetermined reference line.

2. The method according to claim 1 , wherein before extracting:
a refinement process for optimizing size of a gradient and smoothness is performed on the line segment combination.

3. The method according to claim 2 , wherein the refinement process is performed by constructing and solving a balanced energy function for optimizing both of size of the gradient and smoothness.

4. The method according to claim 2 , wherein the refinement process is performed when a discontinuity larger than a predetermined number of pixels is present in the line segment combination.

5. The method according to claim 3 , wherein the refinement process is performed when a discontinuity larger than a predetermined number of pixels is present in the line segment combination.

6. The method according to claim 1 , wherein for a text image of a document type, first setting the predetermined threshold, which is smaller, and the rectangular area, which is smaller, to perform an initial combination on the found line segments, and then setting the predetermined threshold, which is bigger, and the rectangular area, which is bigger, to perform a secondary combination on the found line segments.

7. The method according to claim 4 , wherein for two line segments, whose discontinuity is larger than a predetermined pixel value, a connection line of ends of two line segments, subjected to the refinement process, are extracted as a line segment at the discontinuity.

8. The method according to claim 5 , wherein for two line segments, whose discontinuity is larger than a predetermined pixel value, a connection line of ends of two line segments, subjected to the refinement process, are extracted as a line segment at the discontinuity.

9. A device for extracting a distorted straight line from an image, comprising:
a detecting module configured to perform a line segment detection process on the image;
a combining module configured to, for each line segment detected, find adjacent line segments having a low non-collinear degree with a current line segment, so as to combine them together to obtain one or more line segment combinations; and
an extracting module configured to extract the distorted straight line according to the line segment combination,
wherein finding adjacent line segments having a low non-collinear degree with a current line segment further comprises:
defining an adjoining rectangular area along a slope direction of the current line segment; and
searching another line segment whose non-collinear degree with a current line segment is lower than a predetermined threshold as the adjacent line segment within the adjoining rectangular area in accordance with a principle:
the middle first, then both sides; and
from near to far,

wherein the non-collinear degree is measured by calculating a connection angle with the following formula, the less the connection angle, the lower the non-collinear degree:
connection angle=max(abs( a 2− a 1),abs( a 2− a 3)); wherein

a 1 is an angle between the current line segment and a predetermined reference line;
a 2 is an angle between a connection line of the current line segment with the another line segment and the predetermined reference line; and
a 3 is an angle between the another line segment and the predetermined reference line.

10. The device according to claim 9 , further comprising:
a refinement processing module configured to perform a refinement process for optimizing size of a gradient and smoothness on the line segment combination before the extracting module extracts the distorted straight line.

Assignees

• Fujitsu Ltd

Inventors

• Zheng Yun
• Xie Shufu
• Pan Pan
• He Yuan
• Sun Jun

Classifications

• G06V10/243

  by compensating for image skew or non-uniform image deformations · CPC title

• G06V10/247

  by affine transforms, e.g. correction due to perspective effects; Quadrilaterals, e.g. trapezoids · CPC title

• G06V10/457

  by analysing connectivity, e.g. edge linking, connected component analysis or slices · CPC title

• G06T3/0031

  Physics · mapped topic

• G06T2207/20136

  Physics · mapped topic