How to calculate 3d patch from 3d data which have 512 clices?

Hello I have a problem to calculate 3d patches(blocks) from my 3d dataset. I already know that how to calculate 2d patches like 20 by 20 patch from 2d slice. I want to know, how to make 3d patches from 3d data. like (20 by 20 by 5). If someone have a solution in matlab please share. Thanks

Answer 1

[Walter Roberson]

https://www.mathworks.com/help/matlab/ref/patch.html

"patch('Faces',F,'Vertices',V) creates one or more polygons where V specifies vertex values and F defines which vertices to connect. Specifying only unique vertices and their connection matrix can reduce the size of the data when there are many polygons. Specify one vertex per row in V. To create one polygon, specify F as a vector. To create multiple polygons, specify F as a matrix with one row per polygon. Each face does not have to have the same number of vertices. To specify different numbers of vertices, pad F with NaN values."

In my experience, the much harder part is figuring out what to connect between layers. It is not uncommon to have a larger object slice on one layer than on the next, and larger object slices have more points around the perimeter, so you might have more perimeter points on one slice than on the next slice.

[Ahmad]

this is in some other way. i need a proper way to handle my 3d data using 3d patches to reduce the dimensionality.

[Walter Roberson]

Do you mean that you need to create 3D subsets? If so then you can use mat2cell, or just use indexing.

[Ahmad]

This is my method of 2d patch construction. like 19 by 19 or depends on x_axis_size, y_axes_size. but i need the patch like 19 by 19 by 4 or so on. if you know this or understand my point of view please help me. thanks

 function [image_reshape] = patches_dataset(I, x_axis_size,   y_axes_size)

 imSz = size(I);
 patchSz = [x_axis_size y_axes_size];
 xIdxs = [1:patchSz(2):imSz(2) imSz(2)+1];
 yIdxs = [1:patchSz(1):imSz(1) imSz(1)+1];
 patches = cell(length(yIdxs)-1,length(xIdxs)-1);
 for i = 1:length(yIdxs)-1
    Isub = I(yIdxs(i):yIdxs(i+1)-1,:);
    for j = 1:length(xIdxs)-1
        patches{i,j} = Isub(:,xIdxs(j):xIdxs(j+1)-1);

        patches {i, j} = reshape(patches{i,j},[1, x_axis_size *  y_axes_size]);

        patches{i, j} = (patches{i, j})';
    end
end

 [x, y] = size(patches);
 image_size = x*y;

image_reshape = reshape(patches,[1,image_size]); 

end

[Walter Roberson]

Note by the way that

        patches{i,j} = Isub(:,xIdxs(j):xIdxs(j+1)-1);

        patches {i, j} = reshape(patches{i,j},[1, x_axis_size *  y_axes_size]);

        patches{i, j} = (patches{i, j})';

can be expressed more compactly as just

        patches{i,j} = reshape( Isub(:,xIdxs(j):xIdxs(j+1)-1), [], 1);

[Walter Roberson]

xsize = 19; ysize = 19; zsize = 4;

[nrows, ncols, npanes] = size(I);
if mod(nrows,ysize) ~= 0 || mod(ncol,xsize) ~= 0 || mod(npanes, zsize) ~= 0
   error('Image size [%d, %d, %d] is not evenly divisible into blocks of size [%d, %d, %d]', nrows, ncols, npanes, ysize, xsize, zsize );
end

patches = mat2cell( I, ysize * ones(1, nrows/ysize), xsize * ones(1, ncol/xsize), zsize * ones(1, npanes / zsize) );
patches = reshape( cellfun(@(V) reshape(V,[],1), patches, 'Uniform', 0), 1, [] );