Draw a rectangle over an object using Matlab

I have an image which contains a coin image. Shape of the coin may be rectangle, square, circle, oval and etc. I want to draw a rectangle over the开发者_Go百科 coin and segment the coin from it's background. I can't give x or y values of the rectangle, since coin may be in anywhere in the image. Does anyone know how to do this?

---

My answer below assumes that you have already identified the coin and you have a cleaned image (binary would be nice) that you can work with.

coin=load('penny.mat'); %#load matlab's stock image
img=zeros(256,256);
img(65:192,65:192)=coin.P;%# this is an approximation to the sort of image that I think you have

Now we need the extents of the image in order to know the size of the bounding rectangle. Since the array is non-zero where there's an image and zero elsewhere, the following gives the length of the sides.

sideX=sum(sum(img,1)>0);
sideY=sum(sum(img,2)>0);

Find the centroid of the image using kmeans.

[indX,indY]=ind2sub(size(img),find(img(:)>0));
[~,centroid]=kmeans([indX,indY],1);

Now finally overlay the rectangle on top of the image.

imagesc(img);colormap(gray);hold on
rectangle('Position',([centroid,sideX,sideY]-[sideX,sideY,0,0]/2),'EdgeColor','w');hold off

Result:

If you have a noisy image (i.e., it's not uniformly zero outside the image, then you'd have to set a threshold to find the bounding box)

---

Start by reading the MathWorks advice on pattern recognition. Exactly how you solve the problem depends on many things, for example

• what else is in the image apart from a coin?

• are all the coins the same colour or different colours?

• do you have a training set of images that with identified coin locations?