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

alvarado: @CHG: The methods of ltiGeometricTransform were still in the old interface style. Now they return a boolean. Furthermore, the apply methods for matrices now provide an optional template parameter, which specifies the interpolation functor that shoud be used. This was not implemented as functor parameter due to efficiency issues: as a virtual member, the interpolation would take about 15% more time as a direct instance.

alvarado: @ADD: Two methods added: the first one areaX2() (area times two) compute twice the area of the polygon. Why twice? Because for integer types twice the area is also an integer, and computing on the same type is a little bit more efficient than casting to a floating point type, that in case of interger types would only keep track of a 1/2 term. The second method clockwise() uses areaX2() to decide if the polygon turns clockwise or not.

sltilib: Overload of template methods not supported, so we rename the new one in tApply and tRotate and tScale

alvarado: File name in header fixed.

sltilib: Some type information errors fixed.

alvarado: Bases classes for interpolators are not template classes any more.

alvarado: Initial import. Release LTI-Lib Lithium (v. 1.9.0)

alvarado: Detection of out-of-bounds access fixed.

ltilib: doerfler: moved to math from imgProc
new interpolator hierarchy

alvarado: @ADD: lti::matrix<T> has now two new inline methods with an efficient implementation (implying more storage) to get the index of the last row and last column (lastRow() and lastColumn()), which is consistent with the method lastIdx() of the vector. This was necessary to speed up some algorithms that require this information at the lowest possible cost. All places in other data structures and functors of the LTI-Lib which got this information with rows()-1 or columns()-1 have being updated to the new method. This can speed up a little bit some algorithms like the medianFilter.