Any ideas how I can perform segmentation on this image?

I have tried a range of methods such as clustering and thresholding to seperate the white circle from the brown ring without any luck.

Please could someone help?

Thanks!

[Image Analyst]

Have you tried the color thresholder app on the Apps tab of the tool ribbon? Have you tried findcircles()?

Can you get better focus and lighting and remove the pipette?

[Pranaya Kansakar]

This is from a stack of images from someone else's data, so i cannot remove the pipette

I've tried flatfield correction - it works nicely to remove the shading from the bottom left corner but not the shadow in the white circle.

I've tried imfind circles, but as i said before, these images are from a stack, so some images do not have a circle but rather an elipse. Is there an option to find elipses or circuloid objects instead? Can i manipulate the hough circle transform to achieve this?

I've tried color thresholding - it's goood but patchy (see image) and i have a scalibility issue applying this to the whole stack of images

[Image Analyst]

Use bwconvhull. Do you want the actual boundary, following all the little nooks and crannies, or do you want the best-fit perfect circle?

[Pranaya Kansakar]

Ideally id like to find the area of the white circle

Also,i think using bwconvhull requires binarization of the image.

I'm stuck at that stage as binarization does a terrible job .

Answer 1

[Ryan Comeau]

Hey, i've read through the comments, might have a binarization solution for you. The imbinarize function is very elaborate, take a good look through the documentation. Here is what i've done in the past to get the best binarization results.

image=imread('path/to/image')
image=rescale(image,0,1); %matlab loves 0-1 range for binarize don't ask me why lol
BW_im=imbinarize(image);

If this doesn't work, try a variable sensitivity, adaptive binarization, and create some validation conditions for yourself:

is_done = 0;
sensitivity= 0.1; %tune this to whatever you want 
i=1;
while is_done == 0
    BW(i).image = imbinarize(I,'adaptive','Sensitivity',sensitivity);
    i=i+1;
    sensitivity=sensitivity+0.1;%tune this to what you want.
    if sensitivity >= 1 
        is_done == 1
    end
end

%you then place some validation criterion on you binarization and take which one is the best

Hope this helps,

RC

[Pranaya Kansakar]

Hey Ryan

Thanks for the comments!

I'll give them a go!