Hough transform for circles

A Hough transform function for detecting circles, optimised to allow search over radii.

Description

The Hough transform may be used to detect circular shapes in images, after binarisation, for example by an edge detector. Often, functions to do this operation require the radius of the circle to be specified.

The function circle_hough allows a range of radii to be specified, so that the radius does not need to be known exactly in advance. It is likely to be faster than calling a standard function repeatedly for different radii.

Multiple circles may be detected by finding peaks in the 3D accumulator array which is output. A function, circle_houghpeaks, is provided for this. A demonstration of the two function is included as a script, circle_houghdemo.

The zip file includes an efficient and accurate implementation of circle approximation, without gaps, on an integer grid.

Acknowledgements

The author wishes to acknowledge the following in the creation of this submission:
Hough Transform for circle detection

• Required Products: Image Processing Toolbox
• MATLAB release: MATLAB 7.10 (2010a)

[Truc Phan]

Great job! Thanks for sharing

[Alireza Saberi]

 

[Alireza Saberi]

Hello David,
It should be nice fast work. Here I have following Qs:
1- by running demo file on my LAB system i can see the following error
??? Undefined command/function 'validateattributes'.

Error in ==> circle_hough at 41
validateattributes(rrange, {'double'}, {'real' 'positive' 'vector'});

Error in ==> circle_houghdemo at 26
h = circle_hough(e, radii, 'same', 'normalise');

2- I have an algorithm for fatigue detection of a driver/pilot
the final steps, I have binarized image around pupil and even if the eye is open, i want and i need a good hough transform detect partials of circle of pupil, Can it do that or circle should be perfect for detection?

[David Young]

Alireza Saberi: As this page shows, the code was built using Matlab 2010a. You should not give a low rating just because you have an old version of Matlab. You can delete calls to validateattributes (though you will lose error checking) and it may then run on your system.

On your second point: this isn't the place to discuss a specific application, and I suggest post the question on the newsgroup.

[Alireza Saberi]

 

[Steve Stefanou]

 

[Masih Soltany]

Hi

i want to use your code for my project on " eye detection "

i test but it had many miss place circle detection.

how can i optimize your code ?

thanks

[David Young]

Mash: Sorry, but I can't help. There could be many different reasons why this technique does not work for your project, and you may need to investigate other methods, as well as experimenting to find good parameter values for the edge detection and the HT itself. You might get more detailed advice by posting a request (with example images) on the newsgroup.

[TaoGe ?]

Hello,May Good Job.
But to me,an error found:??? Error: File: circle_houghpeaks.m Line: 113 Column: 11
Expression or statement is incorrect--possibly unbalanced (, {, or [.

How to deal with it?
Thx~

[David Young]

Hello TaoGe, I think you may be using an old version of Matlab that does not allow the use of ~ for anonymous variables. Please try replacing the ~ in line 113 of the file with a variable name, like this:

        [unused, ind] = sort(h(peakind), 'descend');

[Arefeh boroomand]

 

[Tristan]

Awesome Code David, Thanks Alot.
The patch you just mentioned for older versions of matlab worked perfect for me in R2007b

[Nathan Orloff]

I tried several of the different Hough Circle detection algorithms. None of them were as easy to use or as robust as this one. It was also very fast in comparison to the editors pick of the week. Even the script circlepoints turns out to be faster than a standard trick in drawing a circle in matlab. My only suggestion is that you be a little nicer to the newbs David... :)

[Ivan]

wow, thanks a lot for the code David, helped with my project a lot!

just one question,

for example, if i were to set npeaks to 5, this will result in 5 circles being detected. however, i would only like one of the circle.

hence, is there a parameter i can set to eliminate the other circles? or do i have to add on to your code?

thank you so much!

[David Young]

Hi Ivan, thanks for the comment. If you set npeaks to 1, you will get the circle with the strongest evidence. If that is not the right circle, then you will need additional code to select the circle you want, using some other selection criteria, such as radius or position.

[Fer]