Using linspace outside of field data areas?

I have data like this one (data: irregular in xyz, no data in the white area):

and would like to resample it using linspace. How can I force linspace to do this within the data range and set outside values to NaN? Otherwise the result will be like this one:

Thanks a lot for each hint,

Harald

Answer 1

[Walter Roberson]

Use scatteredInterpolant() with the Extrapolation Method set to 'none'

[Harald von der Osten]

thanks a lot for this hint. It works nearly perfect. I just wonder why it doesn't work in the upper area and why the image is somehow mirrowed...:

I used this code:

F = scatteredInterpolant(x,y,z,'linear','none')

xv=linspace(min(x,[],'all'),max(x,[],'all'),(max(x)-min(x))/0.1);
yv=linspace(min(y,[],'all'),max(y,[],'all'),(max(y)-min(y))/0.1);

zg = F({xv,yv});
imagesc(zg); set(gca,'YDir','normal'); axis equal; colormap(flipud(gray));

[Harald von der Osten]

okay, second problem can be solved using:

imagesc(flipud(rot90(zg)))

[Walter Roberson]

Remember that in MATLAB, x is columns (second dimension) and y is rows (first dimension)

[Harald von der Osten]

oh yes, you are right. I didn't remember that. Thanks for pointing this out.

By the way, I still have some problems with

F = scatteredInterpolant(x,y,z,'nearest','none')

For me, it doesn't work perfect. So, is there an alternative method for doing this? For eample, this is a more complicated example:

or should I start a new queston with this?

Again, thanks a lot for your effords,

Harald

[Walter Roberson]

Could you explain more about "it doesn't work perfect" ? What does the uninterpolated version look like?

[Harald von der Osten]

this is the uninterpolated version:

[Walter Roberson]

Could you confirm that the data you use to plot the above is scattered?

And is it correct that you want to interpolate only within "connected" regions? So the "islands" should be done separately, and the "holes" are not to be filled? And the area that looks like it might be a wafer, the white gap should not be filled?

What about the small white areas around the black dots?

[Harald von der Osten]

yes, I can confirm that the data used are scattered. Here is a small sample:

583672.2,5417531.46,0,3.158504376141652,2,0,0,0.1
583672.6800000001,5417531.32,0,5.40855022159389,2,0,1,0.1
583673.16,5417531.19,0,5.839143662601507,2,0,2,0.1
583673.64,5417531.05,0,4.55544778040343,2,0,3,0.1
583674.13,5417530.92,0,6.341357944431953,2,0,4,0.1
583674.61,5417530.79,0,5.714415007217156,2,0,5,0.1
583675.09,5417530.65,0,3.359512393820803,2,0,6,0.1
583675.5699999999,5417530.52,0,2.854370950622065,2,0,7,0.1
583676.05,5417530.38,0,0.8373368253752029,2,0,8,0.1
583676.53,5417530.25,0,0.4138810632575255,2,0,9,0.1
583672.23,5417531.55,0,2.911780594093502,2,0,0,0.2
583672.71,5417531.42,0,4.778725366540664,2,0,1,0.2
583673.1899999999,5417531.29,0,5.314470696520858,2,0,2,0.2
583673.67,5417531.15,0,5.562580691332191,2,0,3,0.2
583674.15,5417531.02,0,6.09445558454945,2,0,4,0.2
583674.63,5417530.88,0,5.189610485588558,2,0,5,0.2
583675.12,5417530.75,0,3.382544976731074,2,0,6,0.2
583675.6,5417530.61,0,2.725916915564379,2,0,7,0.2
583676.08,5417530.48,0,1.22253892156351,2,0,8,0.2
583676.5600000001,5417530.34,0,0.598961386329446,2,0,9,0.2
583672.25,5417531.65,0,3.125056812045332,0,0,0,0.3
583672.73,5417531.52,0,4.950900511487444,0,0,1,0.3
583673.22,5417531.38,0,5.317797730440216,0,0,2,0.3
583673.7,5417531.25,0,6.680713602260954,0,0,3,0.3
583674.1800000001,5417531.11,0,6.068553224666836,0,0,4,0.3
583674.66,5417530.98,0,4.333805963959941,0,0,5,0.3

Yes, I want to interpolate only within connected regions, the white regions should be ignored, white gaps should not be filled. The small white areas around black dots are important (part of the geomagnetic anomalies) and should be not ignored.