Always just try the straightforward solution first--
nMax=10;
[~,ix]=sort(c(:),1,'descend');
[i,j]=ind2sub(size(c),ix(1:nMax)); % keep top nMax, return locations in array
You can do things like keep values with X% of maximum in array to shrink the population drastically before sorting, etc., if the full-sized array turns out to be too time-consuming, but I'd just go "dead-ahead" until was shown to be a real bottleneck...
