Because you didn't write the array indices inside the loop; you filled each array with a constant each pass thru the loop; the last value found will be the value remaining for all elements.
idx_u = nan(m,1); idx_o = idx_u; % fill w/ NaN initially; will overwrite finite locations
for r = 1:m
if any(gradMag(:,r))
ix=find(gradMag(:,r)); % get all locations in one call
idx_o(i)=min(ix); % I'm guessing but you can test that is as
idx_u(i)=max(ix); % fast or faster than two calls to FIND()
end
end
X=1:m; % is just the index vector; no need for inside loop
