Mushroom detection in a photo

Hi, I'm a complete beginner in MATLAB and I'm trying to detect the mushrooms in this photos so I can measure them, etc.

I have tried several algorithms like Hough Circle and detecting and identifying round objects but no luck...

Could someone give me some help please?

[Walter Roberson]

There are clearly Mycelium in the picture, and those effectively have no individual identity and so cannot be counted. You have not really distinguished between those and other commonly identified parts of mushrooms. As you refer to "round", it sounds like you might be wanting to measure the fruiting bodies, which is indeed one of the meanings for "mushroom". Due to the multiple meanings, one could meaningfully but confusingly say "I am interested in the mushroom part of the mushrooms".

Anyhow, please be more specific about what you need to detect.

A lot of the fruiting bodies in the picture are rather small and tucked away and counting them would be difficult.

Answer 1

[Matthew Eicholtz]

When you say you have tried Hough circles, do you mean imfindcircles?

If not, try using that and tweak the input parameters (you may need to increase Sensitivity and decrease EdgeThreshold). You will probably need to call the function multiple times with varying radius ranges to span the set of mushroom sizes in your image.

With that being said, if the Hough transform approach does not work well, it is probably because your image does not have strong edges. An alternative approach that may be worth trying is Laplacian-of-Gaussian (LoG) filters.

h = fspecial('log',hsize,sigma);
J = imfilter(I,h,'replicate','conv');

where I is your image, J is your response map, and (hsize,sigma) are the filter parameters. Detecting circular blobs amounts to finding peaks in the response map.

Here is an online tutorial that may be helpful: Blob Detection by Scale-Space Filtering

Answer 2

[Image Analyst]

Look at Steve's blog where he did a similar thing to count pears: http://www.mathworks.com/help/images/examples/marker-controlled-watershed-segmentation.html I think it's directly applicable.

Also check out http://blogs.mathworks.com/steve/2013/11/19/watershed-transform-question-from-tech-support/