LTI-Lib (C++ Computer Vision Library) Code

The configure file must have mode 7xx

alvarado: @CHG: For the lti::colorModelSelector the computation of the PCA covariance and mean values are now more efficient.

alvarado: @FIX: The lti::matrix couldn't copy a submatrix into the original matrix, if the submatrix shared the memory with the original one. Now this is allowed. Please note that the matrix class still does not track for references, so if there are other matrices sharing the memory, they still will be invalid after removing the "mother" matrix.

alvarado: @FIX: Memory leak with the static Look-Up-Tables of lti::toUCam fixed.

doerfler: revert last change and improve docu

alvarado: A few changes to make it more usable.

alvarado: Members of the parameters transposeU and transposeV moved to the upper class.

alvarado: @FIX: The lti::fastEigensystem functor called the LAPACK function with a too short eigenvalues vector. This leads to an improper memory access that sooner or later was followed by a segmentation fault. Thanks to the wonderful VALGRIND tool, this was not so extremely difficult to find! Well, at least Valgrind said the wrong access was made within the LAPACK function. Difficult was to find which of the 20 arguments was wrong. Many documentation in the code was added to simplify this.

alvarado: @CHG: In lti::draw, the methods related with drawSymbols() are now public.

alvarado: @CHG: The jacobi eigensolution functor now has also a parameters "dimensions" to make it compatible with the lti::fastEigenSystem. Specifying a lower number of dimensions doesn't make this functor any faster although. Maybe there is out there any volunteers to implement a new faster eigensystem functor?