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

alvarado: @ADD: The new data structure lti::priorityQueue is simpler alternative to the std::priority_queue, with the main difference that an identifier for the internal elements is kept valid as soon as the elements exists, no matter of changes in its position within the queue. This allows for example the efficient implementation of algorithms to merge graph nodes that changes the edges values dynamically.

alvarado: @ADD: The lti::regionPolygonizer gets polygonal representations for the regions in a labeled mask. It can ensure vertices for the polygons at those points where three regions are adjacent.

alvarado: @ADD: The lti::multiclassNormalizedCuts functor implements the k-class normalized cuts algorithms used to optimally split a graph into k disconnected subgraphs proposed by Stella Yu in her Ph.D thesis. It requires as input the affinity matrix, which encodes the similarity between two nodes and returns a labels vector, which contains the label assignments for each node.

sltilib: Now compiles also with MS Visual C++

alvarado: Doxygen warnings fixed.

schnurr: added missing include limits

alvarado: @ADD: The edge detection algorithm of Gevers and Stokman has been implemented (lti::colorEdgesGS), which is capable to classify the edges in shadow, highlight or material edges. Is not as fast as lti::cannyEdges for color images, but the categorization of edges can be useful in those applications where the algorithm works.

sltilib: small MVC bug

alvarado: Towards compatibility with Visual .NET 2003

doerfler: @ADD: DBScan is a graph-based clustering algorithm. The number of clusters is dependent on a threshold and clusters are not neccessarily compact.