Counting just coins in this image !...

Hi

I tried below code to count coins in image and it's work fine :

function ret = CountCoins(img)
 subplot(2,2,1);
 imshow(img);
 subplot(2,2,2);
 imgBW = im2bw(img);
 imshow(imgBW);
 subplot(2,2,3);
 imhist(img);
 subplot(2,2,4);
 imgZ = zeros(size(img));
 imgZ(img > 100) = 1;
 imshow(imgZ);
 ret = round(sum(imgBW(:)) / 2100);
 imgConn = bwconncomp(imgZ);
 ret = imgConn.NumObjects;
end

I want just count coins no anything else.

Look...

With above function i can count coins in this image :

Result is 10.(Correct)

But result of this image is wrong;

Result is 12.(Wrong)

because 2 shapes are not coins!

_______________________________________

My question:

How can i count "Just" coins on that image?

[Walter Roberson]

Define "coins" in a way that allows them to be distinguished with a picture of the resolution you show.

Refer to: http://www.joelscoins.com/oops.htm and http://izismile.com/2010/09/01/the_most_unusual_coins_25_pics.html

Answer 1

[Image Analyst]

Did you see my image segmentation tutorial: http://www.mathworks.com/matlabcentral/fileexchange/25157-image-segmentation-tutorial-blobsdemo? I do exactly that, plus a bunch of other useful things. I find 5 cent and 10 cent pieces based on their area. What you want to do is to adapt that to look at something else, like a combination of MeanIntensity and circularity (the ratio of perimeter^2 over (4*pi*area)), or you might want to use the standard deviation in the blob. Then use ismember() to extract out just the blobs that you want. My demo shows you how to do all that, you just need to adapt it slightly, like I said.

[Hamed]

Thank you for your good answer. Sorry i'm very beginner. Please lead me and suggest me some articles if need.

[Image Analyst]

I thought I did that. My tutorial has a comment for virtually every line so that a beginner like yourself should be able to follow it. That's why I wrote it. Let me know of any confusing parts that are not commented well and I'll try to explain them better.

[Walter Roberson]

I'm not sure how you are going to match the guitar coins and distinguish them from guitar jewelry...

[Image Analyst]

Get your binary image like this:

binaryImage = originalImage > 100 & originalImage < 250;
binaryImage = bwareaopen(binaryImage, 500);

It assumes those non-coin shapes will have an intensity of 250 or brighter.