Loop through 3D bins to count numbers of points of each color within each separate bin

Hi all,

I apologize if this is a trivial question; I have mostly only worked with 2D data sets before. I have a large data set consisting of xyz coordinates (locations of electrode readings on a heart), each of which has a 4th-dimensional value assigned to it (representing an activation time reading). I have assigned different time intervals to different colors, resulting in the creation of a color map of activation time. I now want to divide the xyz coordinates in space into a 3D grid, and then analyze each grid section/cube separately to determine how many different colors it contains. I have written some code that uses the 'discretize' function to create bins out of the separate x,y, and z vectors. I am now attempting to analyze through each interval/cube using for loops. For some reason, however, my code is not working the way that I want. Does anyone have any idea how I can define each cube/voxel created by the discretize functions separately, as well as how I can determine the numbers of colors (rgb_time contains the values of the colors assigned to each xyz coordinate) within each cube? Thank you very much!

Answer 1

[Matt J]

If you have followed the advice in your previous thread,

https://www.mathworks.com/matlabcentral/answers/1938929-trying-to-divide-a-3d-scatter-plot-into-grid-to-count-number-of-colored-points-in-each-grid-cell?s_tid=srchtitle

then you should now have a list of voxel IDs for every rgb time. So, you could now do,

counts = splitapply(@(z) height(unique(z,'rows')), RGB, ID(:))

Here, I assume RGB is an Nx3 array of color data.

[AN]

Hi Matt,

Thank you so much for your advice. I think where I am having trouble is in the creation of voxel IDs.. Is there a way to use discretize with a 3 dimensional array rather than using discretize on each vector separately? I can see inputting an array but am not sure what to input for the edges, as the x/y/z vectors all have different start and endpoint values.

[AN]

I have been able to use the following code to create separate x bin IDs, y bin IDs, and z bin IDs for each rgb time:

N = 10;

[xbins,xedges] = discretize(x_points,N);

[ybins,yedges] = discretize(y_points,N);

[zbins,zedges] = discretize(z_points,N);

[Matt J]

Should be easy to find the voxel IDs with findgroups:

voxelIDs=findgroups(xbins,ybins,zbins)

[AN]

I can't thank you enough! I actually just used this code after creating the voxel IDs:

counts = splitapply(@numel, rgb_time, voxelIDs(:));

This seems to create a vector counting the total number of points in each voxel. I did try using your code

counts = splitapply(@(z) height(unique(z,'rows')), RGB, ID(:)) to count how many unique colors are in each voxel but it returns an error saying

Undefined function 'height' for input arguments of type 'double'.

Is 'z' a pre-defined function in MATLAB?

[Matt J]

You should upgrade! But for now, you can do this:

height=@(z) size(z,1);

counts = splitapply(@(z) height(unique(z,'rows')), RGB, ID(:))

Is 'z' a pre-defined function in MATLAB?

No, z is the argument to an anonymous function that we have defined on-the-fly.

[AN]

Amazing! Thank you, you have made my life so much easier. I am working on getting a software upgrade approved currently.

[Matt J]

You are quite welcome, but if your problem is now resolved please Accept-click the answer.