How to find closest values within a matirx

Hi have a matirx

[NAN NAN NAN 7 8 9;
 NAN NAN 5 7 8 7;
 4 8 6 8 3 4;
 4 7 0 1 1 4]

How do I replace the NAN values with the closest values known values. So the matirx will kind of look like

[4 5 7 7 8 9;
 4 8 5 7 8 7;
 4 8 6 8 3 4;
 4 7 0 1 1 4]

[Voss]

A "matrix" with different numbers of elements in different rows is not a matrix.

[NaN NaN NaN 7 8 9;
 NaN NaN 5 7 8;
 4 8 6 8;
 4 7 0 1]

Error using vertcat
Dimensions of arrays being concatenated are not consistent.

So, what do you really have?

[A]

Sorry i think i made a mistake, i didn't count the number of elements properly, I changed it on the question

[Steven Lord]

What are your rules for "closest"? For element (1, 3) you fill in the NaN with the value from (1, 4) rather than the value from (2, 3). But for element (2, 2) you fill in with the value from (3, 2) rather than the value from (2, 3). Why do you choose the element to the side in the first case but the element below in the second?

[A]

Yea it kinda like that logic, like nearest neighbors almost. Becasue for (2,2) the values is close is (3,2) rather than (2,3)

Answer 1

[Matt J]

A=[nan(1,3) 7 8 9;
   nan(1,2) 5 7 8 7;
   4 8 6 8 3 4;
   4 7 0 1 1 4]

A = 4×6

NaN NaN NaN 7 8 9 NaN NaN 5 7 8 7 4 8 6 8 3 4 4 7 0 1 1 4

nanmap=isnan(A);

[~,idx]=bwdist(~nanmap);

A(nanmap)=A(idx(nanmap))

A = 4×6

4 5 5 7 8 9 4 8 5 7 8 7 4 8 6 8 3 4 4 7 0 1 1 4

[A]

That makes more sense, thank you!!

Answer 2

[DGM]

I'm not really sure what you mean by "closest values". Maybe you're talking about inpainting?

A = [NaN NaN NaN 7 8 9;
    NaN NaN 5 7 8 7;
    4 8 6 8 3 4;
    4 7 0 1 1 4]

A = 4×6

NaN NaN NaN 7 8 9 NaN NaN 5 7 8 7 4 8 6 8 3 4 4 7 0 1 1 4

B = regionfill(A,isnan(A))

B = 4×6

5.4778 5.8000 5.9333 7.0000 8.0000 9.0000 5.1556 5.9889 5.0000 7.0000 8.0000 7.0000 4.0000 8.0000 6.0000 8.0000 3.0000 4.0000 4.0000 7.0000 0 1.0000 1.0000 4.0000

If you don't have regionfill() (Image Processing Toolbox), you can always use John's inpaint_nans() on the File Exchange

C = inpaint_nans(A)

You might also be able to use fillmissing() depending on the directional behavior you intend.

[A]

No like creating new values, kind of like subsituing the already known values into the NAN. So kinda like nearest neighbor, if we know what its closes elements are then we subisite them into the NAN

[DGM]

Then Matt's answer might be more what you want