What does a colon followed by comma do to an image?

I don't quite understand what the following code is doing.

numFolder=fullfile('NumberZero/','Zero/');
for i=1:10;
    numName=sprintf('%d.bmp',i);
    image1=imread([numFolder, numName]);
    im1(:,:,i)=image1; % what is this line doing?
end

Answer 1

[Adam]

im1(:,:,i)

means take all elements of your 3d array in the first and 2nd dimensions and only the i'th value in the 3rd dimension.

So basically it turns your data from 3d to 2d in this case by extracting just one slice from the 3rd dimension of a volume.

doc colon

The comma is just the standard syntax for separating the dimensions of your matrix which is why in your case I could easily see that im1 is 3d.

[Recap]

Hi Adam,

Just to make sure that i understand you correctly.

It is turning my input 3d image into a 2d array or my 2d image into a 3d array?

[Adam]

It is turning what starts off as a 3d array into a 2d array by selecting just 1 page/slice of the 3rd dimension.

im1(:,:,1)
im1(:,:,2)
...
im1(:,:,n)

would give you all the 3rd dimension slices of im1 (assuming it has size n in the 3rd dimension) each as 2d arrays.

[Recap]

Great explanation. Thank you

[Image Analyst]

"It is turning what starts off as a 3d array into a 2d array" <=== well, I disagree.

The line

im1(:,:,i)=image1;

takes a 2D image called image1 and sticks it into the i'th slice (plane) of a 3D image called im1. If im1 already has i slices, then it just overwrites the i'th slice. If im1 does not yet have i slices, then this code will append a slice to grow the 3D image in the "Z" direction. So it's turning 2D images stored on your disk into a 3D image. The images1's must be grayscale for this code to work.

[Adam]

Yes, that's true, I forgot about the original posted code and was just looking at the im1(:,:,i) in isolation. Since it is on the left side of an assignment it does behave differently.

[Recap]

@Image Analyst

That makes more sense in what I'm trying to use it for(neural network, number recognition). So it's taking my 2D images of let's say 20 different 0's and growing my 3D image of im1 by appending the 2D image to it everytime?

[Adam]

Yep.

[Recap]

@Adam would you be able to further help me with the full code of this? It's code for training a neural network to recognise numbers but i dont understand how it is working.

[Adam]

I'm not really familiar with neural networks so I'm not sure I would be able to help with that.

[Recap]

The code is there and general comments of what its doing. except that i dont get what its exactly the algorithms are doing because im not familiar with matlab syntax. You can have a look at it from my dropbox if your interested.

https://www.dropbox.com/s/gg0fnxewv72od17/NeuroTrainingNumbers.m?dl=0