Detects multiple disks (coins) in an image using Hough Transform

HOUGHCIRCLES detects multiple disks (coins) in an image using Hough Transform.

Description

HOUGHCIRCLES detects multiple disks (coins) in an image using Hough Transform. The image contains separating, touching, or overlapping disks whose centers may be in or out of the image.  
 
Syntax  
  houghcircles(im, minR, maxR);  
  houghcircles(im, minR, maxR, thresh);  
  houghcircles(im, minR, maxR, thresh, delta);  
  circles = houghcircles(im, minR, maxR);  
  circles = houghcircles(im, minR, maxR, thresh);  
  circles = houghcircles(im, minR, maxR, thresh, delta);  
 
Inputs:  
  - im: input image  
  - minR: minimal radius in pixels  
  - maxR: maximal radius in pixels  
  - thresh (optional): the minimal ratio of the number of detected edge pixels to 0.9 times the calculated circle perimeter (0<thresh<=1, default: 0.33)  
  - delta (optional): the maximal difference between two circles for them to be considered as the same one (default: 12); e.g., c1=(x1 y1 r1), c2=(x2 y2 r2), delta = |x1-x2|+|y1-y2|+|r1-r2|  
 
Output  
  - circles: n-by-4 array of n circles; each circle is represented by (x y r t), where (x y), r, and t are the center coordinate, radius, and ratio of the detected portion to the circle perimeter, respectively. If the output argument is not specified, the original image will be displayed with the detected circles superimposed on it.  

• Required Products: Image Processing Toolbox
• MATLAB release: MATLAB 7.5 (R2007b)

[Sven]

Nice function. Code is well documented and clearly written. 
One suggestion is to follow the example of the peaks() function: if no output argument is given, then create a figure and display the image. If it's used with an output argument, assume the user is embedding the function in their own code, and doesn't want a figure to come up automatically.

[Idillus]

Hi, nice function. I've made some changes that may be interesting. It wold be nice to change the threshold and sigma value for canny edge detection, so, I added a few lines to improve this, changing the location of the definition of the edgeimage 
 
if nargin==3 
     
    thresh = 0.33; % One third of the perimeter 
    delta = 12; % Each element in (x y r) may deviate approx. 4 pixels 
    edgeim = edge(im, 'canny', [0.15 0.2]); 
 
elseif nargin==4 
 
    if ((max(size(canny_th) == 2)) || (max(size(canny_th) == 1)))  
        if max(size(canny_th) == 2) 
            edgeim = edge(im, 'canny', [canny_th(1) canny_th(2)]); 
        end 
        if max(size(canny_th) == 1) 
            edgeim = edge(im, 'canny', canny_th(1)); 
        end 
    end 
    delta = 12; 
     
else (nargin==5) 
 
    if ((max(size(canny_th) == 2)) || (max(size(canny_th) == 1)))  
        if (max(size(canny_th) == 2)) 
            edgeim = edge(im, 'canny', [canny_th(1) canny_th(2)],sigma); 
        end 
        if (max(size(canny_th) == 1)) 
            edgeim = edge(im, 'canny', canny_th(1),sigma); 
        end 
    end 
    delta = 12; 
end 
 

[Idillus]

Sorry, I've made ma mistake 
 
if (nargin >=3 || nargin <= 6) 
     
    if nargin==3 
        thresh = 0.33; % One third of the perimeter 
        delta = 12; % Each element in (x y r) may deviate approx. 4 pixels 
        edgeim = edge(im, 'canny', [0.15 0.2]); 
    end 
     
    if nargin==4 
        edgeim = edge(im, 'canny', [0.15 0.2]); 
        delta = 12; 
    end 
     
    if (nargin==5) 
        edgeim = edge(im, 'canny', canny_th); 
        delta = 12; 
    end 
     
    if (nargin==6) 
       edgeim = edge(im, 'canny', canny_th,sigma); 
       delta=12; 
    end 
    if (nargin == 7) 
         edgeim = edge(im, 'canny', canny_th,sigma); 
    end 
end 
 
if minR<0 || maxR<0 || minR>maxR || thresh<0 || thresh>1 || delta<0 || canny_th <0 || canny_th >1 
  disp('Input conditions: 0<minR, 0<maxR, minR<=maxR, 0<thresh<=1, 0<delta'); 
  return; 
end

[Rezki Al Khairi]

i've try to make with GUI... 
but i cannot place where the instrustion mace be placed... 
can u help me pleasee... 
i really need it..

[sanjay bhattacharya]

i used the houhcircles function. it says 'out of memory' for my test images; but it worked for very small binary images; please help

[Yuan-Liang Tang]

sanjay, 
The most probable cause might be that you invoked the function using inappropriate parameters. The function allocate a 3D matrix of size [size(im,1)+maxR, size(im,2)+maxR, maxR-minR+1], where minR and maxR are the minimal and maximal radii of circles you want to detect, respectively. You may want to check if the size of your input image or the values of minR and maxR are really huge.