Binarizing of a 3D image

I have scanned a pack of fibres using Micro CT machine and now I want to process it using the Matlab. I used below code to processed the .vol file. However, now I want to view this 3D array clearly which means to see the fibres. my expectations are mentioned in below. please correct me if I am wrong.

1)convert the 3D grey scale array in to a binarise image and then start the thresholding.

2)find the optimum thresholding value and add filters accordingly to remove the noise which means to find the air and fibres clearly and then convert back to the 3D grey scale image

Are my expectations are correct? Could you help me to do these things for the 3D array.

Below is the code for my pack of fibers

fid = fopen('E:\FL\fl_D_7mm_6um_1d42s\rfl_D_7mm_6um_1d42s.vol');
voxels = fread(fid, '*uint32');  %read as 32 bit unsigned. I'm assuming they're unsigned
fclose(fid);
voxels = reshape(voxels,[1920,1920,1536]);

%voxels = reshape(voxels, [1920, 1920, 1536]);  %you may need to permute the numbers here
% %and you may need to permute the dimensions afterward

Answer 1

[Image Analyst]

Use a program like Avizo for volume visualization:

https://www.thermofisher.com/us/en/home/industrial/electron-microscopy/electron-microscopy-instruments-workflow-solutions/3d-visualization-analysis-software/avizo-materials-science.html

[Hege]

Thank you Image Analyst. But I have to continue the project using the Matlab as there are big parts that have to be completed using the Matlab. Would you be able to give me some suggesstions please. Appreciate your help!!

[Image Analyst]

Try rgb2gray() and medfilt2(). Perhaps bwareaopen() or imnlnfilt().

There are deep learning noise reduction methods. For example:

This example shows how to train a multiscale context aggregation network (CAN) that is used to approximate an image filtering operation.

The example shows how to train a multiscale CAN network to approximate a bilateral filter, and also provides a pretrained network. If you choose to train the network, use of a CUDA-capable NVIDIA™ GPU with compute capability 3.0 or higher is highly recommended. Use of a GPU requires the Parallel Computing Toolbox™.

[Hege]

Thank you Image Analyst. I am just in a bit inside a problem and appreciate if you could advice me. I tried your suggestion actually. If you note my Micro-CT reading attached code, it is 32 bit and not RGB. 3D array in my workspace is 1920X1920X1536 uint32. Appreciate your advice!

[Image Analyst]

Yes, of course (not sure what I was thinking then). I often create these projection images when doing microct analysis of fibrous materials:

1. maximum intensity projection image
2. top surface image
3. bottom surface
4. thickness image
5. density image
6. basis weight image
7. porosity image

They are done along one axis - top down through the z axis.

[Hege]

Thank you Image analyst.But this code is developed not using the .tif files.it is developed using the .vol file(combination of the .tif files).Actually, what I am trying to do is view the fibers. I uploaded the output of the above code to the volume viewer app in the matlab and fibers are not cleared on it. So, now I am thinking to do any operations for the output of the above code and trying to get the 3D array of the clear image.because 3D array of the clear image is used for lots of analysis in my next works.Appreciate your valuable advices.

[Image Analyst]

I don't know what a .vol format file is, but I'm assuming you have code to read it in and create a 3-D volumetric image.

Not sure exactly what you want but basically you have to get a 3-D image, and segment it into fibers and air. This would typically involve

1. preprocessing the gray scale volumetric image or slices
2. thresholding to produce a binary image
3. post-processing the thresholded image to further clean it up.

Now you'll have two masks - a fiber mask and an air mask. They are not inverses of each other since presumably some voxels were noise - not fibers or air. You can use the volume viewer on the 3-D masks. As you've seen, the volume visualization of MATLAB can't compare to that of Avizo. Basically you're limited to isosurfaces and cutaway views. If you're satisfied with that, then go ahead and use it. You can't do projections or transparency. To get the 7 images I get, I basically have to create them manually, and it's only for one angle of view (top-down).

[Hege]

Thank you very much Image Analyst. we can get .vol file by combining all the .tif files using the Micro-CT. Attached image is the output of above code and visualization through volume Viewer App. Unfortanately fibers are not clear and you can see the elements. Now what I am struggling to get is the clear the image and get the 3D array of the clear image. To make sure we can upload the 3D array of the clear image again to the volume viewer. I am just bit struggling how to clear this 3D array and get the clear image for my next works. Appreciate your advices!

[Image Analyst]

Yes, like I explained, those are the severe limitations of volumeviewer. If you want a real volume visualization program, you'll need to use Avizo or something similar.

Avizo

Kitware VolView

3D Engineering Solutions

[Hege]

Thank you so much Image Analyst.Curently I am thinking How to threshold this 3D array, remove the noise and threshold the air mask and fiber mask.Moreover thinking which filters and optimum thresholding value for this 32 bit 3D array as 3D arrays are new to me.If there is any valuable advices at this stage, it is highly appreciated.

[Image Analyst]

Assuming your fibers are sparse, and your air is predominantly what is there, you would not want to visualize the air since it will just look like a solid opaque volume with only occasional holes in it where the fibers intersect the outer surface. You'd want to visualize the fibers only. You can probably get it with thresholding, and then perhaps with some noise cleanup like bwareaopen().

[Hege]

Thank you Image Analyst. Yes I currently I am looking in to this.

[Hege]

I would appreciate if you could please give me a tip to get the histogram of my grayscale 3D array(1920X1920X1536 uint32). I know for the 2D array it comes by default in '*uint8' using imhist(). but bit confuse in how to get histogram of the 3D array and view it. searched many places.but now in many confusions. your suggestions are appreciated.

[Image Analyst]

Try this:

image3d = randi(255, 100, 100, 100); % Sample data.  Replace with your 3-D array.
histObject = histogram(image3d);
title('Histogram of 3-D Image', 'FontSize', 15);
xlabel('Gray Level', 'FontSize', 15);
ylabel('Count', 'FontSize', 15);
grid on;

[Hege]

Thank you so much for your great help Image Analyst!