% By Philip Torr 2002 % copyright Microsoft Corp. %makes a super quick fit (Torr trick of the trade...) %example given two points we can determine a quick fit to a line as % det [i j k ] % [x y m3] % [x' y' m3] % % where the coefficients of i,j,k are the coefficients of the line %this result generalizes to any dimension allowing for quick fitting of minimal sets without SVD. %for Matlab there is no gain over SVD, but there might be for C++ function result = torr_quick_fit(M) nc = length(M); normM = norm(M); % if (nr ~= nc-1) % error('wrong matrix size in torr_quick_fit'); % end for i = 1:nc A = M(:, [1:i-1 i+1:nc])/normM; result(i) = (-1)^(i+1) * det(A); end result = result'/norm(result);