Line Detection via Standard Hough Transform

Detects lines in binary images via Standard Hough Transform.

Description

Function uses Standard Hough Transform to detect Lines in a binary image.
According to the Hough Transform, each pixel in image space corresponds to a line in Hough space and vise versa.This function uses
polar representation of lines i.e. x*cos(teta)+y*sin(teta)=p to detect
lines in binary image. upper left corner of image is the origin of polar coordinate system
Example :
[pdetect,tetadetect,Accumulator] = houghline(Imbinary,pstep,tetastep,thresh)

Acknowledgements

This file inspired Parabola Detection Using Hough Transform.

• Required Products: Image Processing Toolbox
• MATLAB release: MATLAB 7 (R14)

[Amin Sarafraz]

@Nicolas HUOT, sorry for the late reply.

As mentioned in the help section of the code, the origin of polar coordinate system is top left corner of the image and the rho can not be negative, so I think the code already covers the entire span. Please let me know if I am missing something.

For more open source softwares related to computer vision, please visit http://www.computervisiononline.com/software

[Amin Sarafraz]

@ines, the new version should be corrected for that error. please try it out. You should not have any problem now unless you are using a very large image with very small increment for pdetect that causes the size of pdetect to increase drastically.

Amin

[Pugazhendhi]

The code is very helpful to understand Hough transform. I need a help to find Square using Hough.

[nishant]

 

[Tarun Kothari]

 

[Gijs Huisman]

Unfortunately the bug posted on 13 Jun 2005 is still present and a real one, without this fix one will only find half the hough space. Further the improvement suggested on 30 Aug 2007 by Stijn Helsen does make a significant difference. I understand the original author might not be updating this anymore, it would be good if the community could take owner ship.

[ines]

thank you for this code, but when i compile it i had some errors like :
line 45:elseif thresh < 3
error('threshold must be bigger than or equal to 3')
return;% error :this statement cannot be reached
end

line 76 77 :
pdetect = [];tetadetect = [];
for cnt = 1:length(Potential_p)
if Accumulatortemp(Potential_p(cnt),Potential_teta(cnt)) >= 0
%error: the variable 'pdetect' appears to change size on every lop iteration(same error for 'tetadetect')
pdetect = [pdetect;Potential_p(cnt)];
tetadetect = [tetadetect;Potential_teta(cnt)];
end
end

[Bob]

honestly, your program lacks of efficiency

[collin]

How could we test the accuracy of the detection ? Could you suggest a way? My question applies also when using the standard hough functions.
Thank you

[Ricardo Cabral]

mintemp = min(temp);

Indp = find(temp == mintemp);

Indp = Indp(1);

could be replaced by

[mintemp, indp] = min(temp);

[kiran hidayat]

 

[ale montes]

n/a

[bassam j]

Good program, can be simplified and it can be made faster. But very helpful..

[Junho Lee]

FOr Study

[priya t]

 

[Vijay Kumar]

 

[abbas emadi]

 

[Jia Li]

 

[Stijn Helsen]

My previous comment was not completely correct. pmax shouldn't be p(1), but p(1)-pstep/2

[Stijn Helsen]

Except for the bug (of determining half of the Hough space), the code can be made a lot faster by replacing the "find-method" for determining the p-index to calculation, for example : Indp = round(abs(roi+pmax)/pstep+1);
(where pmax is -p(1)) is a lot faster and as good.

[HAMNETT CGU]

Thanks a lot!!

[Nong Minh Ngoc]

Thanks so much !

[X. King]

It's useful. Good job!

[reg zs]

You are very kindness

[jing zuo]

good

[gaurav goyal]

gud

[zhao tong]

you are very good

[lamya durain]

I am not profesional in mat;ab yet ,but i want to thank you very much for your effort .

[akbar jafary]

 

[ali mot]

 

[Nicolas HUOT]

Hi, I discovered a bug in your file. It was unable to detect Positive sloped lines. I fixed it, tell me if I'm right on this:

function [pdetect,tetadetect,Accumulator] = houghline2(Imbinary,pstep,tetastep,thresh)
%HOUGHLINE - detects lines in a binary image using common computer vision operation known as the Hough Transform.
%
%Comments:
% Function uses Standard Hough Transform to detect Lines in a binary image.
% According to the Hough Transform, each pixel in image space
% corresponds to a line in Hough space and vise versa.This function uses
% polar representation of lines i.e. x*cos(teta)+y*sin(teta)=p to detect
% lines in binary image. upper left corner of image is the origin of polar coordinate
% system.
%
%Usage: [pdetect,tetadetect,Accumulator] = houghline(Imbinary,pstep,tetastep,thresh)
%
%Arguments:
% Imbinary - a binary image. image pixels that have value equal to 1 are
% interested pixels for HOUGHLINE function.
% pstep - interval for radius of lines in polar coordinates.
% tetastep - interval for angle of lines in polar coordinates.
% thresh - a threshold value that determines the minimum number of
% pixels that belong to a line in image space. threshold must
% be bigger than or equal to 3(default).
%
%Returns:
% pdetect - a vactor that contains radius of detected lines in
% polar coordinates system.
% tetadetect - a vector that contains angle of detcted lines in
% polar coordinates system.
% Accumulator - the accumulator array in Hough space.
%
%Originally Written by :
% Amin Sarafraz
% Photogrammetry & Computer Vision Devision
% Geomatics Department,Faculty of Engineering
% University of Tehran,Iran
% [e-mail]
%
%Bug fixed by:
% Nicolas HUOT
% Electrical Engineering and Computer Science Student
% UCLA (Los Angeles) and INPG (Grenoble)
% [e-mail]
%
%May 5,2004 - Original version
%November 24,2004 - Modified version,slightly faster and better performance.
%June 13, 2005 - Positive sloped lines possible (Nicolas HUOT)

if nargin == 3
thresh = 3;
elseif thresh < 3
error('threshold must be bigger than or equal to 3')
return;
end

p = -sqrt((size(Imbinary,1))^2+(size(Imbinary,2))^2):pstep:sqrt((size(Imbinary,1))^2+(size(Imbinary,2))^2);
teta = 0:tetastep:180-tetastep;

%Creating the accumulator:
Accumulator = zeros(length(p),length(teta)); %creating the Accumulator
[yIndex xIndex] = find(Imbinary); %extract the points from the image
for cnt = 1:size(xIndex) %looping on the list of points of the image
Indteta = 0;
for tetai = teta*pi/180 %looping on the list of possible teta (in radians)
Indteta = Indteta+1;
roi = xIndex(cnt)*cos(tetai)+yIndex(cnt)*sin(tetai); %ro computed
if roi >= p(1) & roi <= p(end) %acceptable ro are filtered
temp = abs(roi-p); %computing the distances between the roi and the list of acceptable values
mintemp = min(temp); %picking the minimum of it
Indp = find(temp == mintemp); %finding the indexes of this value
Indp = Indp(1); %picking the first one
Accumulator(Indp,Indteta) = Accumulator(Indp,Indteta)+1; %hop, increasing the coresponding point in the Accu
end
end %end loop tetai
end %end loop image points

% Finding local maxima in Accumulator
AccumulatorbinaryMax = imregionalmax(Accumulator); %creating the array of regional max
[Potential_p Potential_teta] = find(AccumulatorbinaryMax == 1); %picking up the maximums
Accumulatortemp = Accumulator - thresh; %leveling the accu by the threshold
pdetect = [];tetadetect = [];
for cnt = 1:length(Potential_p) %looping on the possible ro
if Accumulatortemp(Potential_p(cnt),Potential_teta(cnt)) >= 0 %if the minimum threshold is reached
pdetect = [pdetect;Potential_p(cnt)]; %append the new confirmed values
tetadetect = [tetadetect;Potential_teta(cnt)];
end
end

% Calculation of detected lines parameters(Radius & Angle).
pdetect = (pdetect - round(sqrt((size(Imbinary,1))^2+(size(Imbinary,2))^2)/pstep)) * pstep;
tetadetect = tetadetect *tetastep - tetastep;

[Morten Engelsmann]

Easily understandable, readily useable

[r zq]

It is a good work.

[Ravi Lakkundi]

We are trying to detect a line in an image, we applied this hough transform., Now we got the accumulator point, please tell us how will we remap that accumulator points back to image space.