Rev | Line | |
---|
[37] | 1 | %normu Normalize image points to be used for LS estimation. |
---|
| 2 | % A = normu(u) finds normalization matrix A so that mean(A*u)=0 and |
---|
| 3 | % mean(||A*u||)=sqrt(2). (see Hartley: In Defence of 8-point Algorithm, |
---|
| 4 | % ICCV`95). |
---|
| 5 | % |
---|
| 6 | % Parameters: |
---|
| 7 | % u ... Size (2,N) or (3,N). Points to normalize. |
---|
| 8 | % A ... Size (3,3). Normalization matrix. |
---|
| 9 | |
---|
| 10 | function A = normu(u) |
---|
| 11 | |
---|
| 12 | if size(u,1)==3, u = p2e(u); end |
---|
| 13 | |
---|
| 14 | m = mean(u')'; % <=> mean (u,2) |
---|
| 15 | u = u - m*ones(1,size(u,2)); |
---|
| 16 | distu = sqrt(sum(u.*u)); |
---|
| 17 | r = mean(distu)/sqrt(2); |
---|
| 18 | if ~r, A = []; return; end % one of degenarate configurations |
---|
| 19 | A = diag([1/r 1/r 1]); |
---|
| 20 | A(1:2,3) = -m/r; |
---|
Note: See
TracBrowser
for help on using the repository browser.