Find first row/column with at least one finite element

Hello :)

I am trying to crop an array to represent it with surf().

The region of interest of the image is always centered and the rest are NaNs, so there woulb be always regions at the top/bot and left/right with rows/columns composed by all NaNs.

I am trying to find the indexes of the first rows/columns containing at least one finite as the array changes depending on the image that is analysed.

This is what I have:

first_row = find(any(isfinite(phases)),1)
last_row = find(any(isfinite(phases)),1, 'last')

first_column = find(any(isfinite(phases)),2)
last_column = find(any(isfinite(phases)),2, 'last')

Then I crop:

surf(phases(first_row:last_row,first_column:last_column));

But I am not having the expected results.

THANK YOU IN ADVANCE.

Gonzalo.

Answer 1

[KSSV]

clc; clear all ;
% demo data 
A = NaN(100) ; 
A(25:75,25:75) = rand(51) ; 
% Get non=nan's index 
idx = ~isnan(A) ; 
% Make X, Y grid for demo 
[X,Y] = meshgrid(1:100,1:100) ; 
% find the first index non-NaN
id = find(idx(:)==1);
[r1,c1] = ind2sub(size(A),id(1)) ; 
[r2,c2] = ind2sub(size(A),id(end)) ;  
%
 pcolor(A)
 shading interp 
 hold on
 plot(X(r1,c1),Y(r1,c1),'*r')
 plot(X(r2,c2),Y(r2,c2),'*r')
 
 

[Gonzalo Molina]

Hello KSSV thank you for your answer,

is there any way to idenitify the rows/columns individually to be used for other purposes?

Actually your answer works for the representation really well but the problem is that I want to crop the array to calculate also some averages later. The representation that I optain with your code comprises the "white" space of the NaNs which is actually somthing that I had before starting to think about the croping.

Thank you.

[KSSV]

You got the first anf last indices right? Now you can pick your region..

[Gonzalo Molina]

Maybe I have explained me in the wrong way, I join here an output with your code with a random geometry.

You can se the two points which correspond to the first/last points from the left/right perspective.

I would like to have aswell the same points at the bottom of the figure, to be able to crop it not only horizontally but vertically.

[KSSV]

You may try this:

Let X, Y,Z be your matrices.

idx = ~isnan(Z) ; 
x = X(idx) ; 
y = Y(idx) ; 
z = Z(idx) ; 
m = 100 ; n = 100 ; 
xi = linspace(min(x),max(x),m) ; 
yi = linspace(min(y),max(y),n) ; 
[X,Y] = meshgrid(xi,yi) ; 
Z = scatteredInterpolant(x,y,z,X,Y) ; 

Or get the bounding box for the region which covers your required.

Pmin = [min(x) min(y)] ; 
Pmax = [max(x) max(y)] ; 

Now you have the bouning box, you can get what you want.