How to replace zero elements with the mean of the closest previous and following values?

for example, given matrix A:

A=[2 0 3 -1 6; -1 -2 3 5 0; 2 3 4 -1 7; 5 1 0 0 4; 0 1 5 0 -2];

I want to obtain this:

A=[2 2.5 3 -1 6; -1 -2 3 5 4; 2 3 4 -1 7; 5 1 2.5 2.5 4; 3 1 5 1.5 -2]

if zero element hasn't previous elements I want the mean between the two following elements

if zero element hasn't following elements I want the mean between the two previous elements

thanks

Answer 1

[Rik]

The code below doesn't average the non-zero above and below for each row, but gets the closest non-zero, with a bias towards the later values.

A=[2 0 3 -1 6; -1 -2 3 5 0; 2 3 4 -1 7; 5 1 0 0 4; 0 1 5 0 -2];
for row=1:size(A,1)
    zeropos=A(row,:)==0;
    if any(zeropos)
        zeropos=find(zeropos);
        non_zero=1:size(A,2);non_zero(zeropos)=[];
        %Use implicit expansion to replicate for multiple zeros, and
        %subtract a small value to force [5 1 0 0 4] to do (1+4)/2
        [~,idx]=sort(abs(non_zero'-zeropos-.1));
        idx=non_zero(idx(1:2,:));
        if size(idx,1)==1,idx=idx';end%flip if only 1 zero
        A(row,zeropos)=(A(row,idx(1,:))+A(row,idx(2,:)))/2;
    end
end

%check
B=[2 2.5 3 -1 6; -1 -2 3 5 4; 2 3 4 -1 7; 5 1 2.5 2.5 4; 3 1 5 1.5 -2];
IsCorrect=~any(abs(A(:)-B(:))>2*eps)

[Zahra Boudreaux]

Thank you for this very good piece of code.