how to interpolate missing data within a 3D cube of data, or for a 2D image with missing slices

Share icon

Hi all,

I am working with a set number 2D images each taken at a different z values (-7 up to +8, steps of 1). I have been able to combine these images into a 3D cube with the z axis being evenly spaced and then used gizmo to view the "surfaces" of the box, no problem.

My problem comes with my second data set, which has missing voltages (was only taken at -7,-5,-3,-1,0,2,4,6,8). I have made this into a 3D cube again, with even spacing again, and have essentially filled in the missing steps with 2D images of NaN values.

My question is: how can I interpolate to fill in this missing data slices?

Currently I am taking a slice in the y,z plane for each given x value, to create a 2D image and am then trying to use Interp2D to fill in the NaN slices but this isn't working (in the second half of the code provided below). This is the image I have attached. How can I interpolate this?
I would then re-populate the 3D cube with the interpolated data.

Thanks in advance for any help!

**********************

Function make_3D_cube()
	
	//make waves to hold the information for the slices and correponding volatges
	string slices="SMP1212_028_kx;temp_neg6;SMP1212_042_kx;temp_neg4;SMP1212_041_kx;temp_neg2;SMP1212_039_kx;SMP1212_037_kx;temp_1;SMP1212_036_kx;temp_3;SMP1212_035_kx;temp_5;SMP1212_034_kx;temp_7;SMP1212_033_kx" 
	Make/O voltages={-7, -6, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, 6, 7, 8} 
 
	String k_scale_str=stringfromlist(0,slices,";")
	wave k_scale = $k_scale_str
	
	//make temp waves to fill in missing voltages with NaN waves
	duplicate/O k_scale, temp_7
	temp_7=nan
	duplicate/O temp_7, temp_neg6, temp_neg4, temp_neg2, temp_1, temp_3, temp_5

	//make 3D cube to poulate
	Make/O/N=(DimSize(k_scale,1), DimSize(k_scale,0), DimSize(voltages,0)) cube_3D
	SetScale/P x, DimOffset(k_scale,1), DimDelta(k_scale,1), "Å^-1", cube_3D
	SetScale/P y, DimOffset(k_scale,0), DimDelta(k_scale,0), "eV", cube_3D
	SetScale/I z, voltages[0], voltages[DimSize(voltages,0)-1], "V", cube_3D 
	cube_3D=Nan
	
	//populate 3D cube from 2D slices
	variable v,i,j
	for(v=0;v nb the code works totally fine up to this point
	//interpolate to fill in the missing data slices  
	//for(i=0;i
Before you start interpolating your data it may be useful to wonder if interpolation is necessary. If you only want to compare your results graphically, e.g., using Gizmo, you may not even need to interpolate.

If you do determine that you need to interpolate, it seems to me that your example data could be interpolated simply by averaging the two adjacent Voltage columns. Interp2d() is not going to be helpful here because you have NaNs in the immediate neighborhood of each interpolated value. My advice is to use a simple MatrixOP command for every NaN column (say colC) so that:
MatrixOP/O interpolatedCol=0.5*(col(temp_2d,colC-1)+col(temp_2d,colC+1))

If your voltage values are in rows, not columns then replace the col() above with row().

A.G.
WaveMetrics, Inc.

October 25, 2017 at 08:26 am - Permalink