replace missing value in a matrix using intropolated values

Dear All,

I have a matrix A as below: A =

    1.1093   -0.7697    1.1006   -0.6156    0.4882   -0.8045    0.1049
       NaN       NaN       NaN       NaN       NaN       NaN       NaN
       NaN       NaN       NaN       NaN       NaN       NaN       NaN
   -1.2141    1.1174   -1.4916    0.8886    1.4193   -0.2437   -0.6669
   -1.1135   -1.0891   -0.7423   -0.7648    0.2916    0.2157    0.1873
   -0.0068       NaN   -1.0616   -1.4023    0.1978   -1.1658   -0.0825
    1.5326    0.5525       NaN   -1.4224    1.5877   -1.1480       NaN

I want to replace the NaN with the values generated from interpolation which can take care of both the interpolation for rows and column. So far I have in mind is take interpolation for row get value a_row; then do it for column get a_column. In the end replace nan with the mean value for the two. Can anyone help me about this? Also, I have problems with specifying y for the function : output(a)= interp1(A,y,'linear');

if I don't know y how should I specify here?

thanks a lot in advance

Answer 1

[Andrei Bobrov]

one way

t = ~isnan(A);
[x,y] = find(t);
F = TriScatteredInterp(x,y,A(t));
[ii,jj] = ndgrid(1:size(A,1),1:size(A,2));
out = F(ii,jj);

ADD variant

out(end) = mean(cellfun(@(x)interp1(1:2,x(1:2),3,'linear','extrap'),...
                       {out(end,end-[2 1 0])', out(end-[2 1 0],end),...
                                out(end - (size(out,1)+1)*(2:-1:0))'}));

[Skirt Zhang]

Hi Bobrov, thanks for your reply. But now I still have NaNs in the output which are located in the end of the matrix. Do you have any idea to solve this problem

[Andrei Bobrov]

Hi Skirt! See ADD part in my answer.