segmentation of irregular shapes

Hi !

I'm doing a school project on whiskers growth and my goal is from SEM images of different whiskers growth steps being able to detect and quantify whiskers and hillocks on that image. So, I thought about matlab segmentation but the issue is that whiskers do not have particular shapes at all neither regular height ( bit more regular for hillocks but still doesn't looks like a perfect half-sphere). So I was wondering if anyone has an idea of a system or a tool that I can use to solve my issue ?

Thanks :P

Answer 1

[KALYAN ACHARJYA]

For segmentation, the ROI object may have any shape, the shape doesnot matter. Segmentation process generally depends on type of image and how much the region of interest is differ from rest part of the image.

I request you to go the basic of segmentation and apply the fundamentatal concepts on your images, we can say or hint on the segmentation method, which may work on your images, only after seeing the images.

Look for basic like thresholding, edge detection refer Golzalez Book of Digital Image Processing using Matlab.

[whiskers]

Hi,

Thank you for you answer as you can see not an absolut fan of matlab :P but I did try automatic segmentation as well as some basic functions such as imopen etc.. Didn't have any sucess with that.

[KALYAN ACHARJYA]

imopen related to morphological operationnormally use after getting the segmentated image, which may further require improvement in results. Can you share image, attach in paper clip.

[whiskers]

Here is on the image I have

[KALYAN ACHARJYA]

Its time consuming, you can try with multiple ways, and check the result accordingly

[T,EM]=graythresh(grayImage);
result3=imbinarize(grayImage,T);
figure, imshow(result3);
title('Otsu method');
%% Basic Thresholding
thresholdValue=150; % You can change this threshold value too
result2=grayImage<=thresholdValue;
%% remove the those pixel area having more that 200 pixels. change accordingly
result2=bwareaopen(result2,200);
figure, imshow(result2);
title('Manual Threshold Value');

Try accordingly, there are numereous segmentation methd, which method suit for the image, you have to look out, this is just for example

[whiskers]

thank you for all you answers and your help !

[KALYAN ACHARJYA]

@whiskers welcome