2D Mask of a 3D image

Hey there,

I'm rather new to using Matlab and my boss has asked me to analyze some SEM images of several membranes to compare the porosity and pore size of various membranes. The images I was given to analyze are 3D black and white images that I'd like to import into Matlab to further analyze. Ideally I'd like to be able to utilize the threshold of the image to identify the depth and thus the volume of the membrane being studied to further characterize the pores but that could be further down the road. My main question is what is the best way to create a 2D mask of the surface layer/cross-section of the image in Matlab? Also, I'd be more than happy to email a sample image if that will help clarify anything. Thanks again.

Sample Photo: https://dl.dropbox.com/u/95356349/3-009-HT-1%20T4.jpg

[Image Analyst]

Upload your image to http://snag.gy

or another site mentioned in http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

Answer 1

[Image Analyst]

Like you said, use the threshold to identify regions of interest:

binaryImage = oneSlice > thresholdValue;

Then call do further processing or call regionprops().

[Andy]

Is there anyway to further automate the process or will it be best for me to simply manipulate the threshold value until I obtain my desired result?

[Image Analyst]

It might be able to be decided automatically if you can figure out some algorithm by looking at the histogram. I don't even know if thresholding is the way to go - you suggested that, not me. Can you upload a few slice images?

[Andy]

Sorry for the delay on my response Image Analyst. I've been trying to effectively splice the image into various layers but have run into quite a bit of trouble doing so which I'm assuming is mostly user error.My main goal is simply to be able to characterize and compare various membranes by their pore distribution and porosity. I have done something similar in the past in ImageJ by adjusting the threshold but as you stated above this method might not be the most suited to accomplish what I desire. From the sample image posted above, do you know of any better way to be able to determine either the pore size or porosity? Thank you again for your help.

[Image Analyst]

It's virtually impossible to automatically determine pore size from an image like that. I don't even know what's a pore in those images. You're best off trying an different method than image analysis to characterize the sample, like one based on air flow or something.

[Andy]

Would there be anyway to indicate any form of depth of the image based of intensity values. I agree with your previous statement and am kind of at a loss for what useful image analysis I could accomplish with these images but as a PhD student I'd like to exhaust all of my resources before telling my PI that something is virtually impossible.

[Image Analyst]

You can say that all pixels of a certain gray level are at the same depth, if that helps you. Not sure if it's true though. Ask your friendly SEM operator for advice and interpretation.

[Andy]

Is there a way to remove any part of the image that doesn't corresponds to a specific gray level?

[Image Analyst]

Define remove. An image has to stay rectangular, though you can set pixels to certain colors, like black.

[Andy]

Could matlab change all the pixels below or above a certain gray level to black thus in a sense creating a mask of the image?

[Image Analyst]

Yes.

imageArray(imageArray<thresholdValue) = 0; % Set to black.

[Andy]

Thank you very much. I'm extremely appreciative for all of your feedback and guidance. Just to clarify because I'm away from a machine with Matlab at the moment. Is imageArray a function of matlab or is that simply notation that I would replace with my image array being manipulated? For instance would the following notation be correct?

I=imread('filepath'); % load image & set threshold
thresholdValue=.90

imageArray(I<thresholdValue)=0; % Set to black

[Image Analyst]

No, not quite - you still have imageArray in there. imageArray is the name of your variable - like the (poorly-named) I in your code. If you have a typical uint8 image, the value would go between 0 and 255. So the code would be:

grayImage = imread(filepath); % Load image 
thresholdValue= 128; % Set threshold value to whatever you want
% Get map of where image is less than the threshold
binaryImage = grayImage < thresholdValue;
% Set pixels meeting threshold criteria to black (zeros).
grayImage (binaryImage) = 0; % Set to black

For the above, I inserted an extra step of creating a binary image to let you see where the image is above or below the threshold. You can imshow(binaryImage) if you want to see what pixels are affected.

[Andy]

Thank you very much.