[ VIGRA Homepage | Function Index | Class Index | Namespaces | File List | Main Page ]

graph_algorithms.hxx
1/************************************************************************/
2/* */
3/* Copyright 2014 by Thorsten Beier and Ullrich Koethe */
4/* */
5/* This file is part of the VIGRA computer vision library. */
6/* The VIGRA Website is */
7/* http://hci.iwr.uni-heidelberg.de/vigra/ */
8/* Please direct questions, bug reports, and contributions to */
9/* ullrich.koethe@iwr.uni-heidelberg.de or */
10/* vigra@informatik.uni-hamburg.de */
11/* */
12/* Permission is hereby granted, free of charge, to any person */
13/* obtaining a copy of this software and associated documentation */
14/* files (the "Software"), to deal in the Software without */
15/* restriction, including without limitation the rights to use, */
16/* copy, modify, merge, publish, distribute, sublicense, and/or */
17/* sell copies of the Software, and to permit persons to whom the */
18/* Software is furnished to do so, subject to the following */
19/* conditions: */
20/* */
21/* The above copyright notice and this permission notice shall be */
22/* included in all copies or substantial portions of the */
23/* Software. */
24/* */
25/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND */
26/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES */
27/* OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND */
28/* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT */
29/* HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, */
30/* WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING */
31/* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR */
32/* OTHER DEALINGS IN THE SOFTWARE. */
33/* */
34/************************************************************************/
35
36/**
37 * This header provides definitions of graph-related algorithms
38 */
39
40#ifndef VIGRA_GRAPH_ALGORITHMS_HXX
41#define VIGRA_GRAPH_ALGORITHMS_HXX
42
43/*std*/
44#include <algorithm>
45#include <vector>
46#include <functional>
47#include <set>
48#include <iomanip>
49
50/*vigra*/
51#include "graphs.hxx"
52#include "graph_generalization.hxx"
53#include "multi_gridgraph.hxx"
54#include "priority_queue.hxx"
55#include "union_find.hxx"
56#include "adjacency_list_graph.hxx"
57#include "graph_maps.hxx"
58
59#include "timing.hxx"
60//#include "openmp_helper.hxx"
61
62
63#include "functorexpression.hxx"
64#include "array_vector.hxx"
65
66namespace vigra{
67
68/** \addtogroup GraphDataStructures
69*/
70//@{
71
72 namespace detail_graph_algorithms{
73 template <class GRAPH_MAP,class COMPERATOR>
74 struct GraphItemCompare
75 {
76
77 GraphItemCompare(const GRAPH_MAP & map,const COMPERATOR & comperator)
78 : map_(map),
79 comperator_(comperator){
80
81 }
82
83 template<class KEY>
84 bool operator()(const KEY & a, const KEY & b) const{
85 return comperator_(map_[a],map_[b]);
86 }
87
88 const GRAPH_MAP & map_;
89 const COMPERATOR & comperator_;
90 };
91 } // namespace detail_graph_algorithms
92
93 /// \brief get a vector of Edge descriptors
94 ///
95 /// Sort the Edge descriptors given weights
96 /// and a comperator
97 template<class GRAPH,class WEIGHTS,class COMPERATOR>
99 const GRAPH & g,
100 const WEIGHTS & weights,
101 const COMPERATOR & comperator,
102 std::vector<typename GRAPH::Edge> & sortedEdges
103 ){
104 sortedEdges.resize(g.edgeNum());
105 size_t c=0;
106 for(typename GRAPH::EdgeIt e(g);e!=lemon::INVALID;++e){
107 sortedEdges[c]=*e;
108 ++c;
109 }
110 detail_graph_algorithms::GraphItemCompare<WEIGHTS,COMPERATOR> edgeComperator(weights,comperator);
112 }
113
114
115 /// \brief copy a lemon node map
116 template<class G,class A,class B>
117 void copyNodeMap(const G & g,const A & a ,B & b){
118 typename G::NodeIt iter(g);
119 while(iter!=lemon::INVALID){
120 b[*iter]=a[*iter];
121 ++iter;
122 }
123
124 }
125 /// \brief copy a lemon edge map
126 template<class G,class A,class B>
127 void copyEdgeMap(const G & g,const A & a ,B & b){
128 typename G::EdgeIt iter(g);
129 while(iter!=lemon::INVALID){
130 b[*iter]=a[*iter];
131 ++iter;
132 }
133 }
134 /// \brief fill a lemon node map
135 template<class G,class A,class T>
136 void fillNodeMap(const G & g, A & a, const T & value){
137 typename G::NodeIt iter(g);
138 while(iter!=lemon::INVALID){
139 a[*iter]=value;
140 ++iter;
141 }
142 }
143 /// \brief fill a lemon edge map
144 template<class G,class A,class T>
145 void fillEdgeMap(const G & g,A & a ,const T & value){
146 typename G::EdgeIt iter(g);
147 while(iter!=lemon::INVALID){
148 a[*iter]=value;
149 ++iter;
150 }
151 }
152
153 /// \brief make a region adjacency graph from a graph and labels w.r.t. that graph
154 ///
155 /// \param graphIn : input graph
156 /// \param labels : labels w.r.t. graphIn
157 /// \param[out] rag : region adjacency graph
158 /// \param[out] affiliatedEdges : a vector of edges of graphIn for each edge in rag
159 /// \param ignoreLabel : optional label to ignore (default: -1 means no label will be ignored)
160 ///
161 template<
162 class GRAPH_IN,
163 class GRAPH_IN_NODE_LABEL_MAP
164 >
169 typename AdjacencyListGraph:: template EdgeMap< std::vector<typename GRAPH_IN::Edge> > & affiliatedEdges,
170 const Int64 ignoreLabel=-1
171 ){
173 typedef typename GraphMapTypeTraits<GRAPH_IN_NODE_LABEL_MAP>::Value LabelType;
174 typedef GRAPH_IN GraphIn;
176
177 typedef typename GraphIn::Edge EdgeGraphIn;
178 typedef typename GraphIn::NodeIt NodeItGraphIn;
179 typedef typename GraphIn::EdgeIt EdgeItGraphIn;
180 typedef typename GraphOut::Edge EdgeGraphOut;
181
182
183 for(NodeItGraphIn iter(graphIn);iter!=lemon::INVALID;++iter){
184 const LabelType l=labels[*iter];
185 if(ignoreLabel==-1 || static_cast<Int64>(l)!=ignoreLabel)
186 rag.addNode(l);
187 }
188
189 for(EdgeItGraphIn e(graphIn);e!=lemon::INVALID;++e){
190 const EdgeGraphIn edge(*e);
191 const LabelType lu = labels[graphIn.u(edge)];
192 const LabelType lv = labels[graphIn.v(edge)];
193 if( lu!=lv && ( ignoreLabel==-1 || (static_cast<Int64>(lu)!=ignoreLabel && static_cast<Int64>(lv)!=ignoreLabel) ) ){
194 // if there is an edge between lu and lv no new edge will be added
195 rag.addEdge( rag.nodeFromId(lu),rag.nodeFromId(lv));
196 }
197 }
198
199 //SET UP HYPEREDGES
200 affiliatedEdges.assign(rag);
201 for(EdgeItGraphIn e(graphIn);e!=lemon::INVALID;++e){
202 const EdgeGraphIn edge(*e);
203 const LabelType lu = labels[graphIn.u(edge)];
204 const LabelType lv = labels[graphIn.v(edge)];
205 //std::cout<<"edge between ?? "<<lu<<" "<<lv<<"\n";
206 if( lu!=lv && ( ignoreLabel==-1 || (static_cast<Int64>(lu)!=ignoreLabel && static_cast<Int64>(lv)!=ignoreLabel) ) ){
207 //std::cout<<"find edge between "<<lu<<" "<<lv<<"\n";
208 EdgeGraphOut ragEdge= rag.findEdge(rag.nodeFromId(lu),rag.nodeFromId(lv));
209 //std::cout<<"invalid?"<<bool(ragEdge==lemon::INVALID)<<" id "<<rag.id(ragEdge)<<"\n";
210 affiliatedEdges[ragEdge].push_back(edge);
211 //std::cout<<"write done\n";
212 }
213 }
214 }
215
216 template<unsigned int DIM, class DTAG, class AFF_EDGES>
217 size_t affiliatedEdgesSerializationSize(
218 const GridGraph<DIM,DTAG> &,
219 const AdjacencyListGraph & rag,
220 const AFF_EDGES & affEdges
221 ){
222 size_t size = 0;
223
224 typedef typename AdjacencyListGraph::EdgeIt EdgeIt;
225 typedef typename AdjacencyListGraph::Edge Edge;
226
227 for(EdgeIt iter(rag); iter!=lemon::INVALID; ++iter){
228 const Edge e(*iter);
229 size+=1;
230 size+=affEdges[e].size()*(DIM+1);
231 }
232 return size;
233 }
234
235 template<class OUT_ITER, unsigned int DIM, class DTAG, class AFF_EDGES>
236 void serializeAffiliatedEdges(
237 const GridGraph<DIM,DTAG> &,
238 const AdjacencyListGraph & rag,
239 const AFF_EDGES & affEdges,
240 OUT_ITER outIter
241 ){
242
243 typedef typename AdjacencyListGraph::EdgeIt EdgeIt;
244 typedef typename AdjacencyListGraph::Edge Edge;
245 typedef typename GridGraph<DIM,DTAG>::Edge GEdge;
246
247 for(EdgeIt iter(rag); iter!=lemon::INVALID; ++iter){
248
249 const Edge edge = *iter;
250 const size_t numAffEdge = affEdges[edge].size();
251 *outIter = numAffEdge; ++outIter;
252
253 for(size_t i=0; i<numAffEdge; ++i){
254 const GEdge gEdge = affEdges[edge][i];
255 for(size_t d=0; d<DIM+1; ++d){
256 *outIter = gEdge[d]; ++outIter;
257 }
258 }
259 }
260 }
261
262 template<class IN_ITER, unsigned int DIM, class DTAG, class AFF_EDGES>
263 void deserializeAffiliatedEdges(
264 const GridGraph<DIM,DTAG> &,
265 const AdjacencyListGraph & rag,
266 AFF_EDGES & affEdges,
267 IN_ITER begin,
268 IN_ITER
269 ){
270
271 typedef typename AdjacencyListGraph::EdgeIt EdgeIt;
272 typedef typename AdjacencyListGraph::Edge Edge;
273 typedef typename GridGraph<DIM,DTAG>::Edge GEdge;
274
275 affEdges.assign(rag);
276
277 for(EdgeIt iter(rag); iter!=lemon::INVALID; ++iter){
278
279 const Edge edge = *iter;
280 const size_t numAffEdge = *begin; ++begin;
281
282 for(size_t i=0; i<numAffEdge; ++i){
283 GEdge gEdge;
284 for(size_t d=0; d<DIM+1; ++d){
285 gEdge[d]=*begin; ++begin;
286 }
287 affEdges[edge].push_back(gEdge);
288 }
289 }
290 }
291
292
293
294
295 /// \brief shortest path computer
296 template<class GRAPH,class WEIGHT_TYPE>
298 public:
299 typedef GRAPH Graph;
300
301 typedef typename Graph::Node Node;
302 typedef typename Graph::NodeIt NodeIt;
303 typedef typename Graph::Edge Edge;
304 typedef typename Graph::OutArcIt OutArcIt;
305
306 typedef WEIGHT_TYPE WeightType;
308 typedef typename Graph:: template NodeMap<Node> PredecessorsMap;
309 typedef typename Graph:: template NodeMap<WeightType> DistanceMap;
311
312 /// \ brief constructor from graph
313 ShortestPathDijkstra(const Graph & g)
314 : graph_(g),
315 pq_(g.maxNodeId()+1),
316 predMap_(g),
317 distMap_(g)
318 {
319 }
320
321 /// \brief run shortest path given edge weights
322 ///
323 /// \param weights : edge weights encoding the distance between adjacent nodes (must be non-negative)
324 /// \param source : source node where shortest path should start
325 /// \param target : target node where shortest path should stop. If target is not given
326 /// or <tt>INVALID</tt>, the shortest path from source to all reachable nodes is computed
327 /// \param maxDistance : path search is terminated when the path length exceeds <tt>maxDistance</tt>
328 ///
329 /// When a valid \a target is unreachable from \a source (either because the graph is disconnected
330 /// or \a maxDistance is exceeded), it is set to <tt>lemon::INVALID</tt>. In contrast, if \a target
331 /// was <tt>lemon::INVALID</tt> at the beginning, it will always be set to the last node
332 /// visited in the search.
333 template<class WEIGHTS>
334 void run(const WEIGHTS & weights, const Node & source,
335 const Node & target = lemon::INVALID,
336 WeightType maxDistance=NumericTraits<WeightType>::max())
337 {
338 this->initializeMaps(source);
339 runImpl(weights, target, maxDistance);
340 }
341
342 /// \brief run shortest path in a region of interest of a \ref GridGraph.
343 ///
344 /// \param start : first point in the desired ROI.
345 /// \param stop : beyond the last point in the desired ROI (i.e. exclusive)
346 /// \param weights : edge weights encoding the distance between adjacent nodes (must be non-negative)
347 /// \param source : source node where shortest path should start
348 /// \param target : target node where shortest path should stop. If target is not given
349 /// or <tt>INVALID</tt>, the shortest path from source to all reachable nodes is computed
350 /// \param maxDistance : path search is terminated when the path length exceeds <tt>maxDistance</tt>
351 ///
352 /// This version of <tt>run()</tt> restricts the path search to the ROI <tt>[start, stop)</tt> and only
353 /// works for instances of \ref GridGraph. Otherwise, it is identical to the standard <tt>run()</tt>
354 /// function.
355 template<class WEIGHTS>
356 void run(Node const & start, Node const & stop,
357 const WEIGHTS & weights, const Node & source,
358 const Node & target = lemon::INVALID,
359 WeightType maxDistance=NumericTraits<WeightType>::max())
360 {
361 vigra_precondition(allLessEqual(start, source) && allLess(source, stop),
362 "ShortestPathDijkstra::run(): source is not within ROI");
363 vigra_precondition(target == lemon::INVALID ||
364 (allLessEqual(start, target) && allLess(target, stop)),
365 "ShortestPathDijkstra::run(): target is not within ROI");
366 this->initializeMaps(source, start, stop);
367 runImpl(weights, target, maxDistance);
368 }
369
370 /// \brief run shortest path again with given edge weights
371 ///
372 /// This only differs from standard <tt>run()</tt> by initialization: Instead of resetting
373 /// the entire graph, this only resets the nodes that have been visited in the
374 /// previous run, i.e. the contents of the array <tt>discoveryOrder()</tt>.
375 /// This will be much faster if only a small fraction of the nodes has to be reset.
376 template<class WEIGHTS>
377 void reRun(const WEIGHTS & weights, const Node & source,
378 const Node & target = lemon::INVALID,
379 WeightType maxDistance=NumericTraits<WeightType>::max())
380 {
381 this->reInitializeMaps(source);
382 runImpl(weights, target, maxDistance);
383 }
384
385 /// \brief run shortest path with given edge weights from multiple sources.
386 ///
387 /// This is otherwise identical to standard <tt>run()</tt>, except that
388 /// <tt>source()</tt> returns <tt>lemon::INVALID</tt> after path search finishes.
389 template<class WEIGHTS, class ITER>
390 void
392 const Node & target = lemon::INVALID,
393 WeightType maxDistance=NumericTraits<WeightType>::max())
394 {
395 this->initializeMapsMultiSource(source_begin, source_end);
396 runImpl(weights, target, maxDistance);
397 }
398
399
400 /// \brief run shortest path with given edge weights from multiple sources.
401 ///
402 /// This is otherwise identical to standard <tt>run()</tt>, except that
403 /// <tt>source()</tt> returns <tt>lemon::INVALID</tt> after path search finishes.
404 template<class EFGE_WEIGHTS,class NODE_WEIGHTS, class ITER>
405 void
407 const EFGE_WEIGHTS & edgeWeights,
411 const Node & target = lemon::INVALID,
412 WeightType maxDistance = NumericTraits<WeightType>::max())
413 {
414 this->initializeMapsMultiSource(source_begin, source_end);
415 runImplWithNodeWeights(edgeWeights, nodeWeights, target, maxDistance);
416 }
417
418 /// \brief get the graph
419 const Graph & graph()const{
420 return graph_;
421 }
422 /// \brief get the source node
423 const Node & source()const{
424 return source_;
425 }
426 /// \brief get the target node
427 const Node & target()const{
428 return target_;
429 }
430
431 /// \brief check if explicit target is given
432 bool hasTarget()const{
433 return target_!=lemon::INVALID;
434 }
435
436 /// \brief get an array with all visited nodes, sorted by distance from source
438 return discoveryOrder_;
439 }
440
441 /// \brief get the predecessors node map (after a call of run)
442 const PredecessorsMap & predecessors()const{
443 return predMap_;
444 }
445
446 /// \brief get the distances node map (after a call of run)
447 const DistanceMap & distances()const{
448 return distMap_;
449 }
450
451 /// \brief get the distance to a rarget node (after a call of run)
452 WeightType distance(const Node & target)const{
453 return distMap_[target];
454 }
455
456
457 private:
458
459 template<class WEIGHTS>
460 void runImpl(const WEIGHTS & weights,
461 const Node & target = lemon::INVALID,
462 WeightType maxDistance=NumericTraits<WeightType>::max())
463 {
465 this->runImplWithNodeWeights(weights,zeroNodeMap, target, maxDistance);
466 }
467
468
469 template<class EDGE_WEIGHTS, class NODE_WEIGHTS>
470 void runImplWithNodeWeights(
471 const EDGE_WEIGHTS & edgeWeights,
472 const NODE_WEIGHTS & nodeWeights,
473 const Node & target = lemon::INVALID,
474 WeightType maxDistance=NumericTraits<WeightType>::max())
475 {
476 target_ = lemon::INVALID;
477 while(!pq_.empty() ){ //&& !finished){
478 const Node topNode(graph_.nodeFromId(pq_.top()));
479 if(distMap_[topNode] > maxDistance)
480 break; // distance threshold exceeded
481 pq_.pop();
482 discoveryOrder_.push_back(topNode);
483 if(topNode == target)
484 break;
485 // loop over all neigbours
486 for(OutArcIt outArcIt(graph_,topNode);outArcIt!=lemon::INVALID;++outArcIt){
487 const Node otherNode = graph_.target(*outArcIt);
488 const size_t otherNodeId = graph_.id(otherNode);
489 const WeightType otherNodeWeight = nodeWeights[otherNode];
490 if(pq_.contains(otherNodeId)){
491 const Edge edge(*outArcIt);
492 const WeightType currentDist = distMap_[otherNode];
493 const WeightType alternativeDist = distMap_[topNode]+edgeWeights[edge]+otherNodeWeight;
494 if(alternativeDist<currentDist){
495 pq_.push(otherNodeId,alternativeDist);
496 distMap_[otherNode]=alternativeDist;
497 predMap_[otherNode]=topNode;
498 }
499 }
500 else if(predMap_[otherNode]==lemon::INVALID){
501 const Edge edge(*outArcIt);
502 const WeightType initialDist = distMap_[topNode]+edgeWeights[edge]+otherNodeWeight;
503 if(initialDist<=maxDistance)
504 {
505 pq_.push(otherNodeId,initialDist);
506 distMap_[otherNode]=initialDist;
507 predMap_[otherNode]=topNode;
508 }
509 }
510 }
511 }
512 while(!pq_.empty() ){
513 const Node topNode(graph_.nodeFromId(pq_.top()));
514 predMap_[topNode]=lemon::INVALID;
515 pq_.pop();
516 }
517 if(target == lemon::INVALID || discoveryOrder_.back() == target)
518 target_ = discoveryOrder_.back(); // Means that target was reached. If, to the contrary, target
519 // was unreachable within maxDistance, target_ remains INVALID.
520 }
521
522 void initializeMaps(Node const & source){
523 for(NodeIt n(graph_); n!=lemon::INVALID; ++n){
524 const Node node(*n);
525 predMap_[node]=lemon::INVALID;
526 }
527 distMap_[source]=static_cast<WeightType>(0.0);
528 predMap_[source]=source;
529 discoveryOrder_.clear();
530 pq_.push(graph_.id(source),0.0);
531 source_=source;
532 }
533
534 void initializeMaps(Node const & source,
535 Node const & start, Node const & stop)
536 {
537 Node left_border = min(start, Node(1)),
538 right_border = min(predMap_.shape()-stop, Node(1)),
539 DONT_TOUCH = Node(lemon::INVALID) - Node(1);
540
541 initMultiArrayBorder(predMap_.subarray(start-left_border, stop+right_border),
542 left_border, right_border, DONT_TOUCH);
543 predMap_.subarray(start, stop) = lemon::INVALID;
544 predMap_[source]=source;
545
546 distMap_[source]=static_cast<WeightType>(0.0);
547 discoveryOrder_.clear();
548 pq_.push(graph_.id(source),0.0);
549 source_=source;
550 }
551
552 template <class ITER>
553 void initializeMapsMultiSource(ITER source, ITER source_end){
554 for(NodeIt n(graph_); n!=lemon::INVALID; ++n){
555 const Node node(*n);
556 predMap_[node]=lemon::INVALID;
557 }
558 discoveryOrder_.clear();
559 for( ; source != source_end; ++source)
560 {
561 distMap_[*source]=static_cast<WeightType>(0.0);
562 predMap_[*source]=*source;
563 pq_.push(graph_.id(*source),0.0);
564 }
565 source_=lemon::INVALID;
566 }
567
568 void reInitializeMaps(Node const & source){
569 for(unsigned int n=0; n<discoveryOrder_.size(); ++n){
570 predMap_[discoveryOrder_[n]]=lemon::INVALID;
571 }
572 distMap_[source]=static_cast<WeightType>(0.0);
573 predMap_[source]=source;
574 discoveryOrder_.clear();
575 pq_.push(graph_.id(source),0.0);
576 source_=source;
577 }
578
579 const Graph & graph_;
580 PqType pq_;
581 PredecessorsMap predMap_;
582 DistanceMap distMap_;
583 DiscoveryOrder discoveryOrder_;
584
585 Node source_;
586 Node target_;
587 };
588
589 /// \brief get the length in node units of a path
590 template<class NODE,class PREDECESSORS>
592 const NODE source,
593 const NODE target,
594 const PREDECESSORS & predecessors
595 ){
596 if(predecessors[target]==lemon::INVALID)
597 return 0;
598 else{
599 NODE currentNode = target;
600 size_t length=1;
601 while(currentNode!=source){
602 currentNode=predecessors[currentNode];
603 length+=1;
604 }
605 return length;
606 }
607 }
608
609 /// \brief Astar Shortest path search
610 template<class GRAPH,class WEIGHTS,class PREDECESSORS,class DISTANCE,class HEURSTIC>
612 const GRAPH & graph,
613 const typename GRAPH::Node & source,
614 const typename GRAPH::Node & target,
615 const WEIGHTS & weights,
616 PREDECESSORS & predecessors,
617 DISTANCE & distance,
618 const HEURSTIC & heuristic
619 ){
620
621 typedef GRAPH Graph;
622 typedef typename Graph::Edge Edge;
623 typedef typename Graph::Node Node;
624 typedef typename Graph::NodeIt NodeIt;
625 typedef typename Graph::OutArcIt OutArcIt;
626 typedef typename DISTANCE::value_type DistanceType;
627
628 typename GRAPH:: template NodeMap<bool> closedSet(graph);
630 // initialize
631 for(NodeIt n(graph);n!=lemon::INVALID;++n){
632 const Node node(*n);
633 closedSet[node]=false;
634 distance[node]=std::numeric_limits<DistanceType>::infinity();
635 predecessors[node]=lemon::INVALID;
636 }
637 // distance and estimated distance for start node
638 distance[source]=static_cast<DistanceType>(0.0);
639 estimatedDistanceOpenSet.push(graph.id(source),heuristic(source,target));
640
641 // while any nodes left in openSet
642 while(!estimatedDistanceOpenSet.empty()){
643
644 // get the node with the lpwest estimated distance in the open set
645 const Node current = graph.nodeFromId(estimatedDistanceOpenSet.top());
646
647 // reached target?
648 if(current==target)
649 break;
650
651 // remove current from openSet
652 // add current to closedSet
654 closedSet[current]=true;
655
656 // iterate over neigbours of current
657 for(OutArcIt outArcIt(graph,current);outArcIt!=lemon::INVALID;++outArcIt){
658
659 // get neigbour node and id
660 const Node neighbour = graph.target(*outArcIt);
661 const size_t neighbourId = graph.id(neighbour);
662
663 // if neighbour is not yet in closedSet
664 if(!closedSet[neighbour]){
665
666 // get edge between current and neigbour
667 const Edge edge(*outArcIt);
668
669 // get tentative score
670 const DistanceType tenativeScore = distance[current] + weights[edge];
671
672 // neighbour NOT in openSet OR tentative score better than the current distance
673 if(!estimatedDistanceOpenSet.contains(neighbourId) || tenativeScore < distance[neighbour] ){
674 // set predecessors and distance
675 predecessors[neighbour]=current;
676 distance[neighbour]=tenativeScore;
677
678 // update the estimated cost from neighbour to target
679 // ( and neigbour will be (re)-added to openSet)
681 }
682 }
683 }
684 }
685 }
686
687
688 template<
689 class GRAPH,
690 class EDGE_WEIGHTS,
691 class NODE_WEIGHTS,
692 class SEED_NODE_MAP,
693 class WEIGHT_TYPE
694 >
695 void shortestPathSegmentation(
696 const GRAPH & graph,
697 const EDGE_WEIGHTS & edgeWeights,
698 const NODE_WEIGHTS & nodeWeights,
699 SEED_NODE_MAP & seeds
700 ){
701
702 typedef GRAPH Graph;
703 typedef typename Graph::Node Node;
704 typedef typename Graph::NodeIt NodeIt;
705 typedef WEIGHT_TYPE WeightType;
706
707 // find seeds
708 std::vector<Node> seededNodes;
709 for(NodeIt n(graph);n!=lemon::INVALID;++n){
710 const Node node(*n);
711 // not a seed
712 if(seeds[node]!=0){
713 seededNodes.push_back(node);
714 }
715 }
716
717 // do shortest path
718 typedef ShortestPathDijkstra<Graph, WeightType> Sp;
719 typedef typename Sp::PredecessorsMap PredecessorsMap;
720 Sp sp(graph);
721 sp.runMultiSource(edgeWeights, nodeWeights, seededNodes.begin(), seededNodes.end());
722 const PredecessorsMap & predMap = sp.predecessors();
723 // do the labeling
724 for(NodeIt n(graph);n!=lemon::INVALID;++n){
725 Node node(*n);
726 if(seeds[node]==0){
727 Node pred=predMap[node];
728 while(seeds[pred]==0){
729 pred=predMap[pred];
730 }
731 seeds[node]=seeds[pred];
732 }
733 }
734 }
735
736 namespace detail_watersheds_segmentation{
737
738 struct RawPriorityFunctor{
739 template<class LabelType, class T>
740 T operator()(const LabelType /*label*/,const T priority)const{
741 return priority;
742 }
743 };
744
745
746
747 template<class PRIORITY_TYPE,class LABEL_TYPE>
748 struct CarvingFunctor{
749 CarvingFunctor(const LABEL_TYPE backgroundLabel,
750 const PRIORITY_TYPE & factor,
751 const PRIORITY_TYPE & noPriorBelow
752 )
753 : backgroundLabel_(backgroundLabel),
754 factor_(factor),
755 noPriorBelow_(noPriorBelow){
756 }
757 PRIORITY_TYPE operator()(const LABEL_TYPE label,const PRIORITY_TYPE priority)const{
758 if(priority>=noPriorBelow_)
759 return (label==backgroundLabel_ ? priority*factor_ : priority);
760 else{
761 return priority;
762 }
763 }
764 LABEL_TYPE backgroundLabel_;
765 PRIORITY_TYPE factor_;
766 PRIORITY_TYPE noPriorBelow_;
767 };
768
769
770 template<
771 class GRAPH,
772 class EDGE_WEIGHTS,
773 class SEEDS,
774 class PRIORITY_MANIP_FUNCTOR,
775 class LABELS
776 >
777 void edgeWeightedWatershedsSegmentationImpl(
778 const GRAPH & g,
779 const EDGE_WEIGHTS & edgeWeights,
780 const SEEDS & seeds,
781 PRIORITY_MANIP_FUNCTOR & priorManipFunctor,
782 LABELS & labels
783 ){
784 typedef GRAPH Graph;
785 typedef typename Graph::Edge Edge;
786 typedef typename Graph::Node Node;
787 typedef typename Graph::NodeIt NodeIt;
788 typedef typename Graph::OutArcIt OutArcIt;
789
790 typedef typename EDGE_WEIGHTS::Value WeightType;
791 typedef typename LABELS::Value LabelType;
792 //typedef typename Graph:: template EdgeMap<bool> EdgeBoolMap;
793 typedef PriorityQueue<Edge,WeightType,true> PQ;
794
795 PQ pq;
796 //EdgeBoolMap inPQ(g);
797 copyNodeMap(g,seeds,labels);
798 //fillEdgeMap(g,inPQ,false);
799
800
801 // put edges from nodes with seed on pq
802 for(NodeIt n(g);n!=lemon::INVALID;++n){
803 const Node node(*n);
804 if(labels[node]!=static_cast<LabelType>(0)){
805 for(OutArcIt a(g,node);a!=lemon::INVALID;++a){
806 const Edge edge(*a);
807 const Node neigbour=g.target(*a);
808 //std::cout<<"n- node "<<g.id(neigbour)<<"\n";
809 if(labels[neigbour]==static_cast<LabelType>(0)){
810 const WeightType priority = priorManipFunctor(labels[node],edgeWeights[edge]);
811 pq.push(edge,priority);
812 //inPQ[edge]=true;
813 }
814 }
815 }
816 }
817
818
819 while(!pq.empty()){
820
821 const Edge edge = pq.top();
822 pq.pop();
823
824 const Node u = g.u(edge);
825 const Node v = g.v(edge);
826 const LabelType lU = labels[u];
827 const LabelType lV = labels[v];
828
829
830 if(lU==0 && lV==0){
831 throw std::runtime_error("both have no labels");
832 }
833 else if(lU!=0 && lV!=0){
834 // nothing to do
835 }
836 else{
837
838 const Node unlabeledNode = lU==0 ? u : v;
839 const LabelType label = lU==0 ? lV : lU;
840
841 // assign label to unlabeled node
842 labels[unlabeledNode] = label;
843
844 // iterate over the nodes edges
845 for(OutArcIt a(g,unlabeledNode);a!=lemon::INVALID;++a){
846 const Edge otherEdge(*a);
847 const Node targetNode=g.target(*a);
848 if(labels[targetNode] == 0){
849 //if(inPQ[otherEdge] == false && labels[targetNode] == 0){
850 const WeightType priority = priorManipFunctor(label,edgeWeights[otherEdge]);
851 pq.push(otherEdge,priority);
852 // inPQ[otherEdge]=true;
853 }
854 }
855 }
856
857 }
858
859 }
860
861 } // end namespace detail_watersheds_segmentation
862
863
864 /// \brief edge weighted watersheds Segmentataion
865 ///
866 /// \param g: input graph
867 /// \param edgeWeights : edge weights / edge indicator
868 /// \param seeds : seed must be non empty!
869 /// \param[out] labels : resulting nodeLabeling (not necessarily dense)
870 template<class GRAPH,class EDGE_WEIGHTS,class SEEDS,class LABELS>
872 const GRAPH & g,
874 const SEEDS & seeds,
875 LABELS & labels
876 ){
877 detail_watersheds_segmentation::RawPriorityFunctor fPriority;
878 detail_watersheds_segmentation::edgeWeightedWatershedsSegmentationImpl(g,edgeWeights,seeds,fPriority,labels);
879 }
880
881
882 /// \brief edge weighted watersheds Segmentataion
883 ///
884 /// \param g: input graph
885 /// \param edgeWeights : edge weights / edge indicator
886 /// \param seeds : seed must be non empty!
887 /// \param backgroundLabel : which label is background
888 /// \param backgroundBias : bias for background
889 /// \param noPriorBelow : don't bias the background if edge indicator is below this value
890 /// \param[out] labels : resulting nodeLabeling (not necessarily dense)
891 template<class GRAPH,class EDGE_WEIGHTS,class SEEDS,class LABELS>
893 const GRAPH & g,
895 const SEEDS & seeds,
896 const typename LABELS::Value backgroundLabel,
897 const typename EDGE_WEIGHTS::Value backgroundBias,
898 const typename EDGE_WEIGHTS::Value noPriorBelow,
899 LABELS & labels
900 ){
901 typedef typename EDGE_WEIGHTS::Value WeightType;
902 typedef typename LABELS::Value LabelType;
903 detail_watersheds_segmentation::CarvingFunctor<WeightType,LabelType> fPriority(backgroundLabel,backgroundBias, noPriorBelow);
904 detail_watersheds_segmentation::edgeWeightedWatershedsSegmentationImpl(g,edgeWeights,seeds,fPriority,labels);
905 }
906
907 /// \brief edge weighted watersheds Segmentataion
908 ///
909 /// \param graph: input graph
910 /// \param edgeWeights : edge weights / edge indicator
911 /// \param nodeSizes : size of each node
912 /// \param k : free parameter of felzenszwalb algorithm
913 /// \param[out] nodeLabeling : nodeLabeling (not necessarily dense)
914 /// \param nodeNumStopCond : optional stopping condition
915 template< class GRAPH , class EDGE_WEIGHTS, class NODE_SIZE,class NODE_LABEL_MAP>
917 const GRAPH & graph,
919 const NODE_SIZE & nodeSizes,
920 float k,
922 const int nodeNumStopCond = -1
923 ){
924 typedef GRAPH Graph;
925 typedef typename Graph::Edge Edge;
926 typedef typename Graph::Node Node;
927
928 typedef typename EDGE_WEIGHTS::Value WeightType;
929 typedef typename EDGE_WEIGHTS::Value NodeSizeType;
930 typedef typename Graph:: template NodeMap<WeightType> NodeIntDiffMap;
931 typedef typename Graph:: template NodeMap<NodeSizeType> NodeSizeAccMap;
932
933 // initalize node size map and internal diff map
937 fillNodeMap(graph,internalDiff,static_cast<WeightType>(0.0));
938
939
940
941 // initlaize internal node diff map
942
943 // sort the edges by their weights
944 std::vector<Edge> sortedEdges;
945 std::less<WeightType> comperator;
947
948 // make the ufd
949 UnionFindArray<UInt64> ufdArray(graph.maxNodeId()+1);
950
951
952 size_t nodeNum = graph.nodeNum();
953
954
955 while(true){
956 // iterate over edges is the sorted order
957 for(size_t i=0;i<sortedEdges.size();++i){
958 const Edge e = sortedEdges[i];
959 const size_t rui = ufdArray.findIndex(graph.id(graph.u(e)));
960 const size_t rvi = ufdArray.findIndex(graph.id(graph.v(e)));
961 const Node ru = graph.nodeFromId(rui);
962 const Node rv = graph.nodeFromId(rvi);
963 if(rui!=rvi){
964
965 //check if to merge or not ?
966 const WeightType w = edgeWeights[e];
969 const WeightType tauRu = static_cast<WeightType>(k)/static_cast<WeightType>(sizeRu);
970 const WeightType tauRv = static_cast<WeightType>(k)/static_cast<WeightType>(sizeRv);
971 const WeightType minIntDiff = std::min(internalDiff[ru]+tauRu,internalDiff[rv]+tauRv);
972 if(w<=minIntDiff){
973 // do merge
974 ufdArray.makeUnion(rui,rvi);
975 --nodeNum;
976 // update size and internal difference
977 const size_t newRepId = ufdArray.findIndex(rui);
978 const Node newRepNode = graph.nodeFromId(newRepId);
981 }
982 }
983 if(nodeNumStopCond >= 0 && nodeNum==static_cast<size_t>(nodeNumStopCond)){
984 break;
985 }
986 }
987 if(nodeNumStopCond==-1){
988 break;
989 }
990 else{
991 if(nodeNumStopCond >= 0 && nodeNum>static_cast<size_t>(nodeNumStopCond)){
992 k *= 1.2f;
993 }
994 else{
995 break;
996 }
997 }
998 }
999 ufdArray.makeContiguous();
1000 for(typename GRAPH::NodeIt n(graph);n!=lemon::INVALID;++n){
1001 const Node node(*n);
1002 nodeLabeling[node]=ufdArray.findLabel(graph.id(node));
1003 }
1004 }
1005
1006
1007
1008
1009 namespace detail_graph_smoothing{
1010
1011 template<
1012 class GRAPH,
1013 class NODE_FEATURES_IN,
1014 class EDGE_WEIGHTS,
1015 class WEIGHTS_TO_SMOOTH_FACTOR,
1016 class NODE_FEATURES_OUT
1017 >
1018 void graphSmoothingImpl(
1019 const GRAPH & g,
1020 const NODE_FEATURES_IN & nodeFeaturesIn,
1021 const EDGE_WEIGHTS & edgeWeights,
1022 WEIGHTS_TO_SMOOTH_FACTOR & weightsToSmoothFactor,
1023 NODE_FEATURES_OUT & nodeFeaturesOut
1024
1025 ){
1026 typedef GRAPH Graph;
1027 typedef typename Graph::Edge Edge;
1028 typedef typename Graph::Node Node;
1029 typedef typename Graph::NodeIt NodeIt;
1030 typedef typename Graph::OutArcIt OutArcIt;
1031
1032 typedef typename NODE_FEATURES_IN::Value NodeFeatureInValue;
1033 typedef typename NODE_FEATURES_OUT::Reference NodeFeatureOutRef;
1034 typedef typename EDGE_WEIGHTS::ConstReference SmoothFactorType;
1035
1036
1037 //fillNodeMap(g, nodeFeaturesOut, typename NODE_FEATURES_OUT::value_type(0.0));
1038
1039 for(NodeIt n(g);n!=lemon::INVALID;++n){
1040
1041 const Node node(*n);
1042
1043 NodeFeatureInValue featIn = nodeFeaturesIn[node];
1044 NodeFeatureOutRef featOut = nodeFeaturesOut[node];
1045
1046 featOut=0;
1047 float weightSum = 0.0;
1048 size_t degree = 0;
1049 for(OutArcIt a(g,node);a!=lemon::INVALID;++a){
1050 const Edge edge(*a);
1051 const Node neigbour(g.target(*a));
1052 SmoothFactorType smoothFactor= weightsToSmoothFactor(edgeWeights[edge]);
1053
1054 NodeFeatureInValue neighbourFeat = nodeFeaturesIn[neigbour];
1055 neighbourFeat*=smoothFactor;
1056 if(degree==0)
1057 featOut = neighbourFeat;
1058 else
1059 featOut += neighbourFeat;
1060 weightSum+=smoothFactor;
1061 ++degree;
1062 }
1063 // fixme..set me to right type
1064 featIn*=static_cast<float>(degree);
1065 weightSum+=static_cast<float>(degree);
1066 featOut+=featIn;
1067 featOut/=weightSum;
1068 }
1069 }
1070
1071 template<class T>
1072 struct ExpSmoothFactor{
1073 ExpSmoothFactor(const T lambda,const T edgeThreshold,const T scale)
1074 : lambda_(lambda),
1075 edgeThreshold_(edgeThreshold),
1076 scale_(scale){
1077 }
1078 T operator()(const T weight){
1079 return weight> edgeThreshold_ ? 0 : std::exp(-1.0*lambda_*weight)*scale_;
1080 }
1081 T lambda_;
1082 T edgeThreshold_;
1083 T scale_;
1084 };
1085
1086 } // namespace detail_graph_smoothing
1087
1088
1089 /// \brief smooth node features of a graph
1090 ///
1091 /// \param g : input graph
1092 /// \param nodeFeaturesIn : input node features which should be smoothed
1093 /// \param edgeIndicator : edge indicator to indicate over which edges one should smooth
1094 /// \param lambda : scale edge indicator by lambda before taking negative exponent
1095 /// \param edgeThreshold : edge threshold
1096 /// \param scale : how much smoothing should be applied
1097 /// \param[out] nodeFeaturesOut : smoothed node features
1098 template<class GRAPH, class NODE_FEATURES_IN,class EDGE_INDICATOR,class NODE_FEATURES_OUT>
1100 const GRAPH & g,
1103 const float lambda,
1104 const float edgeThreshold,
1105 const float scale,
1107 ){
1108 detail_graph_smoothing::ExpSmoothFactor<float> functor(lambda,edgeThreshold,scale);
1109 detail_graph_smoothing::graphSmoothingImpl(g,nodeFeaturesIn,edgeIndicator,functor,nodeFeaturesOut);
1110 }
1111
1112 /// \brief smooth node features of a graph
1113 ///
1114 /// \param g : input graph
1115 /// \param nodeFeaturesIn : input node features which should be smoothed
1116 /// \param edgeIndicator : edge indicator to indicate over which edges one should smooth
1117 /// \param lambda : scale edge indicator by lambda before taking negative exponent
1118 /// \param edgeThreshold : edge threshold
1119 /// \param scale : how much smoothing should be applied
1120 /// \param iterations : how often should this algorithm be called recursively
1121 /// \param[out] nodeFeaturesBuffer : preallocated(!) buffer to store node features temp.
1122 /// \param[out] nodeFeaturesOut : smoothed node features
1123 template<class GRAPH, class NODE_FEATURES_IN,class EDGE_INDICATOR,class NODE_FEATURES_OUT>
1125 const GRAPH & g,
1128 const float lambda,
1129 const float edgeThreshold,
1130 const float scale,
1131 size_t iterations,
1134 ){
1135
1136 iterations = std::max(size_t(1),iterations);
1137 // initial run
1139 iterations -=1;
1140
1141 bool outAsIn=true;
1142 for(size_t i=0;i<iterations;++i){
1143 if(outAsIn){
1145 outAsIn=false;
1146 }
1147 else{
1149 outAsIn=true;
1150 }
1151 }
1152 if(!outAsIn){
1154 }
1155 }
1156
1157
1158 template<class GRAPH, class NODE_MAP, class EDGE_MAP>
1159 void nodeGtToEdgeGt(
1160 const GRAPH & g,
1161 const NODE_MAP & nodeGt,
1162 const Int64 ignoreLabel,
1163 EDGE_MAP & edgeGt
1164 ){
1165 typedef typename GRAPH::Node Node;
1166 typedef typename GRAPH::EdgeIt EdgeIt;
1167 typedef typename GRAPH::Edge Edge;
1168
1169 for(EdgeIt edgeIt(g); edgeIt!=lemon::INVALID; ++edgeIt){
1170 const Edge edge(*edgeIt);
1171 const Node u = g.u(edge);
1172 const Node v = g.v(edge);
1173
1174 const Int64 lU = static_cast<Int64>(nodeGt[u]);
1175 const Int64 lV = static_cast<Int64>(nodeGt[v]);
1176
1177 if(ignoreLabel==-1 || (lU!=ignoreLabel || lV!=ignoreLabel)){
1178 edgeGt[edge] = lU == lV ? 0 : 1;
1179 }
1180 else{
1181 edgeGt[edge] = 2;
1182 }
1183 }
1184 }
1185
1186
1187
1188
1189
1190
1191 /// project ground truth from a base graph, to a region adjacency graph.
1192 ///
1193 ///
1194 ///
1195 ///
1196 template<class RAG, class BASE_GRAPH, class BASE_GRAPH_RAG_LABELS,
1197 class BASE_GRAPH_GT, class RAG_GT, class RAG_GT_QT>
1199 const RAG & rag,
1200 const BASE_GRAPH & baseGraph,
1202 const BASE_GRAPH_GT & baseGraphGt,
1203 RAG_GT & ragGt,
1204 RAG_GT_QT &
1205 ){
1206 typedef typename BASE_GRAPH::Node BaseGraphNode;
1207 typedef typename BASE_GRAPH::NodeIt BaseGraphNodeIt;
1208 typedef typename RAG::NodeIt RagNodeIt;
1209 typedef typename RAG::Node RagNode;
1210 typedef typename BASE_GRAPH_RAG_LABELS::Value BaseRagLabelType;
1211 typedef typename BASE_GRAPH_GT::Value GtLabelType;
1212 typedef typename RAG_GT::Value RagGtLabelType;
1213
1214 // overlap map
1215 typedef std::map<GtLabelType,UInt32> MapType;
1216 typedef typename MapType::const_iterator MapIter;
1217 typedef typename RAG:: template NodeMap<MapType> Overlap;
1219
1220 size_t i=0;
1221 //::cout<<"\n";
1222 for(BaseGraphNodeIt baseNodeIter(baseGraph); baseNodeIter!=lemon::INVALID; ++baseNodeIter , ++i ){
1223
1224 //if (i%2000 == 0){
1225 // std::cout<<"\r"<<std::setw(10)<< float(i)/float(baseGraph.nodeNum())<<std::flush;
1226 //}
1227
1228 const BaseGraphNode baseNode = *baseNodeIter;
1229
1230 // get gt label
1231 const GtLabelType gtLabel = baseGraphGt[baseNode];
1232
1233 // get the label which generated rag
1234 // (node mapping from bg-node to rag-node-id)
1236 const RagNode ragNode = rag.nodeFromId(bgRagLabel);
1237
1238 // fill overlap
1239 overlap[ragNode][gtLabel]+=1;
1240 }
1241 //std::cout<<"\n";
1242 // select label with max overlap
1243 for(RagNodeIt ragNodeIter(rag); ragNodeIter!=lemon::INVALID; ++ragNodeIter ){
1244 const RagNode ragNode = *ragNodeIter;
1245 const MapType olMap = overlap[ragNode];
1246 UInt32 olSize=0;
1248 for(MapIter olIter = olMap.begin(); olIter!=olMap.end(); ++olIter ){
1249 if(olIter->second > olSize){
1250 olSize = olIter->second;
1251 bestLabel = olIter->first;
1252 }
1253 }
1255 }
1256 }
1257
1258
1259
1260 /// \brief Find indices of points on the edges
1261 ///
1262 /// \param rag : Region adjacency graph of the labels array
1263 /// \param graph : Graph of labels array
1264 /// \param affiliatedEdges : The affiliated edges of the region adjacency graph
1265 /// \param labelsArray : The label image
1266 /// \param node : The node (of the region adjacency graph), whose edges shall be found
1267 template<class RAGGRAPH, class GRAPH, class RAGEDGES, unsigned int N, class T>
1269 const RAGGRAPH & rag,
1270 const GRAPH & graph,
1271 const RAGEDGES & affiliatedEdges,
1273 const typename RAGGRAPH::Node & node
1274 ){
1275 typedef typename GRAPH::Node Node;
1276 typedef typename GRAPH::Edge Edge;
1277 typedef typename RAGGRAPH::OutArcIt RagOutArcIt;
1278 typedef typename RAGGRAPH::Edge RagEdge;
1280
1281 T nodeLabel = rag.id(node);
1282
1283 // Find edges and write them into a set.
1284 std::set< NodeCoordinate > edgeCoordinates;
1285 for (RagOutArcIt iter(rag, node); iter != lemon::INVALID; ++iter)
1286 {
1287 const RagEdge ragEdge(*iter);
1288 const std::vector<Edge> & affEdges = affiliatedEdges[ragEdge];
1289 for (int i=0; i<affEdges.size(); ++i)
1290 {
1291 Node u = graph.u(affEdges[i]);
1292 Node v = graph.v(affEdges[i]);
1293 T uLabel = labelsArray[u];
1294 T vLabel = labelsArray[v];
1295
1297 if (uLabel == nodeLabel)
1298 {
1299 coords = GraphDescriptorToMultiArrayIndex<GRAPH>::intrinsicNodeCoordinate(graph, u);
1300 }
1301 else if (vLabel == nodeLabel)
1302 {
1303 coords = GraphDescriptorToMultiArrayIndex<GRAPH>::intrinsicNodeCoordinate(graph, v);
1304 }
1305 else
1306 {
1307 vigra_precondition(false, "You should not come to this part of the code.");
1308 }
1309 edgeCoordinates.insert(coords);
1310 }
1311 }
1312
1313 // Fill the return array.
1315 edgePoints.init(0);
1316 int next = 0;
1317 typedef typename std::set< NodeCoordinate >::iterator setIter;
1318 for (setIter iter = edgeCoordinates.begin(); iter!=edgeCoordinates.end(); ++iter)
1319 {
1320 NodeCoordinate coords = *iter;
1321 for (int k=0; k<coords.size(); ++k)
1322 {
1323 edgePoints(next, k) = coords[k];
1324 }
1325 next++;
1326 }
1327 return edgePoints;
1328 }
1329 /// \brief create edge weights from node weights
1330 ///
1331 /// \param g : input graph
1332 /// \param nodeWeights : node property map holding node weights
1333 /// \param[out] edgeWeights : resulting edge weights
1334 /// \param euclidean : if 'true', multiply the computed weights with the Euclidean
1335 /// distance between the edge's end nodes (default: 'false')
1336 /// \param func : binary function that computes the edge weight from the
1337 /// weights of the edge's end nodes (default: take the average)
1338 template<unsigned int N, class DirectedTag,
1339 class NODEMAP, class EDGEMAP, class FUNCTOR>
1340 void
1342 const GridGraph<N, DirectedTag> & g,
1343 const NODEMAP & nodeWeights,
1345 bool euclidean,
1346 FUNCTOR const & func)
1347 {
1348 typedef GridGraph<N, DirectedTag> Graph;
1349 typedef typename Graph::Edge Edge;
1350 typedef typename Graph::EdgeIt EdgeIt;
1351 typedef typename MultiArrayShape<N>::type CoordType;
1352
1353 vigra_precondition(nodeWeights.shape() == g.shape(),
1354 "edgeWeightsFromNodeWeights(): shape mismatch between graph and nodeWeights.");
1355
1356 for (EdgeIt iter(g); iter!=lemon::INVALID; ++iter)
1357 {
1358 const Edge edge(*iter);
1359 const CoordType uCoord(g.u(edge));
1360 const CoordType vCoord(g.v(edge));
1361 if (euclidean)
1362 {
1364 }
1365 else
1366 {
1368 }
1369 }
1370 }
1371
1372 template<unsigned int N, class DirectedTag,
1373 class NODEMAP, class EDGEMAP>
1374 inline void
1376 const GridGraph<N, DirectedTag> & g,
1377 const NODEMAP & nodeWeights,
1378 EDGEMAP & edgeWeights,
1379 bool euclidean=false)
1380 {
1381 using namespace vigra::functor;
1383 }
1384
1385
1386 /// \brief create edge weights from an interpolated image
1387 ///
1388 /// \param g : input graph
1389 /// \param interpolatedImage : interpolated image
1390 /// \param[out] edgeWeights : edge weights
1391 /// \param euclidean : if 'true', multiply the weights with the Euclidean
1392 /// distance between the edge's end nodes (default: 'false')
1393 ///
1394 /// For each edge, the function reads the weight from <tt>interpolatedImage[u+v]</tt>,
1395 /// where <tt>u</tt> and <tt>v</tt> are the coordinates of the edge's end points.
1396 template<unsigned int N, class DirectedTag,
1397 class T, class EDGEMAP>
1398 void
1400 const GridGraph<N, DirectedTag> & g,
1403 bool euclidean = false)
1404 {
1405 typedef GridGraph<N, DirectedTag> Graph;
1406 typedef typename Graph::Edge Edge;
1407 typedef typename Graph::EdgeIt EdgeIt;
1408 typedef typename MultiArrayShape<N>::type CoordType;
1409
1410 vigra_precondition(interpolatedImage.shape() == 2*g.shape()-CoordType(1),
1411 "edgeWeightsFromInterpolatedImage(): interpolated shape must be shape*2-1");
1412
1413 for (EdgeIt iter(g); iter!=lemon::INVALID; ++iter)
1414 {
1415 const Edge edge(*iter);
1416 const CoordType uCoord(g.u(edge));
1417 const CoordType vCoord(g.v(edge));
1418 if (euclidean)
1419 {
1421 }
1422 else
1423 {
1425 }
1426 }
1427 }
1428
1429 template<class GRAPH>
1430 struct ThreeCycle{
1431
1432 typedef typename GRAPH::Node Node;
1433
1434 ThreeCycle(const Node & a, const Node & b, const Node c){
1435 nodes_[0] = a;
1436 nodes_[1] = b;
1437 nodes_[2] = c;
1438 std::sort(nodes_, nodes_+3);
1439 }
1440 bool operator < (const ThreeCycle & other)const{
1441 if(nodes_[0] < other.nodes_[0]){
1442 return true;
1443 }
1444 else if(nodes_[0] == other.nodes_[0]){
1445 if(nodes_[1] < other.nodes_[1]){
1446 return true;
1447 }
1448 else if(nodes_[1] == other.nodes_[1]){
1449 if(nodes_[2] < other.nodes_[2]){
1450 return true;
1451 }
1452 else{
1453 return false;
1454 }
1455 }
1456 else{
1457 return false;
1458 }
1459 }
1460 else{
1461 return false;
1462 }
1463 }
1464
1465 Node nodes_[3];
1466
1467
1468 };
1469
1470
1471 template<class GRAPH>
1472 void find3Cycles(
1473 const GRAPH & g,
1475 ){
1476 typedef typename GRAPH::Node Node;
1477 typedef typename GRAPH::Edge Edge;
1478 typedef typename GRAPH::EdgeIt EdgeIt;
1479 typedef typename GRAPH::OutArcIt OutArcIt;
1480
1481 typedef ThreeCycle<GRAPH> Cycle;
1482
1483 std::set< Cycle > cycles;
1484 typedef typename std::set<Cycle>::const_iterator SetIter;
1485 for (EdgeIt iter(g); iter!=lemon::INVALID; ++iter){
1486 const Edge edge(*iter);
1487 const Node u = g.u(edge);
1488 const Node v = g.v(edge);
1489
1490 // find a node n which is connected to u and v
1491 for(OutArcIt outArcIt(g,u); outArcIt!=lemon::INVALID;++outArcIt){
1492 const Node w = g.target(*outArcIt);
1493 if(w != v){
1494 const Edge e = g.findEdge(w,v);
1495 if(e != lemon::INVALID){
1496 // found cycle
1497 cycles.insert(Cycle(u, v, w));
1498 }
1499 }
1500 }
1501 }
1503 UInt32 i=0;
1504 for(SetIter iter=cycles.begin(); iter!=cycles.end(); ++iter){
1505
1506 const Cycle & c = *iter;
1507 for(size_t j=0;j<3; ++j){
1508 cyclesArray(i)[j] = g.id(c.nodes_[j]);
1509 }
1510 ++i;
1511 }
1512 }
1513
1514 template<class GRAPH>
1515 void find3CyclesEdges(
1516 const GRAPH & g,
1518 ){
1519 typedef typename GRAPH::Node Node;
1520 typedef typename GRAPH::Edge Edge;
1521 typedef typename GRAPH::EdgeIt EdgeIt;
1522 typedef typename GRAPH::OutArcIt OutArcIt;
1523
1524 typedef ThreeCycle<GRAPH> Cycle;
1525
1526 std::set< Cycle > cycles;
1527 typedef typename std::set<Cycle>::const_iterator SetIter;
1528 for (EdgeIt iter(g); iter!=lemon::INVALID; ++iter){
1529 const Edge edge(*iter);
1530 const Node u = g.u(edge);
1531 const Node v = g.v(edge);
1532
1533 // find a node n which is connected to u and v
1534 for(OutArcIt outArcIt(g,u); outArcIt!=lemon::INVALID;++outArcIt){
1535 const Node w = g.target(*outArcIt);
1536 if(w != v){
1537 const Edge e = g.findEdge(w,v);
1538 if(e != lemon::INVALID){
1539 // found cycle
1540 cycles.insert(Cycle(u, v, w));
1541 }
1542 }
1543 }
1544 }
1546 UInt32 i=0;
1547 for(SetIter iter=cycles.begin(); iter!=cycles.end(); ++iter){
1548
1549 const Cycle & c = *iter;
1550 const Node u = c.nodes_[0];
1551 const Node v = c.nodes_[1];
1552 const Node w = c.nodes_[2];
1553
1554 cyclesArray(i)[0] = g.id(g.findEdge(u, v));
1555 cyclesArray(i)[1] = g.id(g.findEdge(u, w));
1556 cyclesArray(i)[2] = g.id(g.findEdge(v, w));
1557 ++i;
1558 }
1559 }
1560
1561//@}
1562
1563} // namespace vigra
1564
1565#endif // VIGRA_GRAPH_ALGORITHMS_HXX
undirected adjacency list graph in the LEMON API
Definition adjacency_list_graph.hxx:228
detail_adjacency_list_graph::ItemIter< GraphType, Edge > EdgeIt
edge iterator
Definition adjacency_list_graph.hxx:253
detail::GenericEdge< index_type > Edge
edge descriptor
Definition adjacency_list_graph.hxx:249
size_type size() const
Definition array_vector.hxx:358
reference back()
Definition array_vector.hxx:321
const_reference top() const
get index with top priority
Definition priority_queue.hxx:490
void pop()
Remove the current top element.
Definition priority_queue.hxx:502
bool empty() const
check if the PQ is empty
Definition priority_queue.hxx:444
bool contains(const int i) const
check if i is an index on the PQ
Definition priority_queue.hxx:459
void push(const value_type i, const priority_type p)
Insert a index with a given priority.
Definition priority_queue.hxx:475
MultiArrayShape< N+1 >::type Edge
The edge descriptor (API: LEMON).
Definition multi_gridgraph.hxx:1592
Main MultiArray class containing the memory management.
Definition multi_array.hxx:2479
Class for a single RGB value.
Definition rgbvalue.hxx:128
shortest path computer
Definition graph_algorithms.hxx:297
WeightType distance(const Node &target) const
get the distance to a rarget node (after a call of run)
Definition graph_algorithms.hxx:452
void runMultiSource(const WEIGHTS &weights, ITER source_begin, ITER source_end, const Node &target=lemon::INVALID, WeightType maxDistance=NumericTraits< WeightType >::max())
run shortest path with given edge weights from multiple sources.
Definition graph_algorithms.hxx:391
void run(Node const &start, Node const &stop, const WEIGHTS &weights, const Node &source, const Node &target=lemon::INVALID, WeightType maxDistance=NumericTraits< WeightType >::max())
run shortest path in a region of interest of a GridGraph.
Definition graph_algorithms.hxx:356
ShortestPathDijkstra(const Graph &g)
\ brief constructor from graph
Definition graph_algorithms.hxx:313
const PredecessorsMap & predecessors() const
get the predecessors node map (after a call of run)
Definition graph_algorithms.hxx:442
const DiscoveryOrder & discoveryOrder() const
get an array with all visited nodes, sorted by distance from source
Definition graph_algorithms.hxx:437
const Graph & graph() const
get the graph
Definition graph_algorithms.hxx:419
const Node & target() const
get the target node
Definition graph_algorithms.hxx:427
const Node & source() const
get the source node
Definition graph_algorithms.hxx:423
void runMultiSource(const EFGE_WEIGHTS &edgeWeights, const NODE_WEIGHTS &nodeWeights, ITER source_begin, ITER source_end, const Node &target=lemon::INVALID, WeightType maxDistance=NumericTraits< WeightType >::max())
run shortest path with given edge weights from multiple sources.
Definition graph_algorithms.hxx:406
bool hasTarget() const
check if explicit target is given
Definition graph_algorithms.hxx:432
void run(const WEIGHTS &weights, const Node &source, const Node &target=lemon::INVALID, WeightType maxDistance=NumericTraits< WeightType >::max())
run shortest path given edge weights
Definition graph_algorithms.hxx:334
const DistanceMap & distances() const
get the distances node map (after a call of run)
Definition graph_algorithms.hxx:447
void reRun(const WEIGHTS &weights, const Node &source, const Node &target=lemon::INVALID, WeightType maxDistance=NumericTraits< WeightType >::max())
run shortest path again with given edge weights
Definition graph_algorithms.hxx:377
void init(Iterator i, Iterator end)
Definition tinyvector.hxx:708
size_type size() const
Definition tinyvector.hxx:913
iterator end()
Definition tinyvector.hxx:864
iterator begin()
Definition tinyvector.hxx:861
Class for fixed size vectors.
Definition tinyvector.hxx:1008
void projectGroundTruth(const RAG &rag, const BASE_GRAPH &baseGraph, const BASE_GRAPH_RAG_LABELS &baseGraphRagLabels, const BASE_GRAPH_GT &baseGraphGt, RAG_GT &ragGt, RAG_GT_QT &)
Definition graph_algorithms.hxx:1198
void copyNodeMap(const G &g, const A &a, B &b)
copy a lemon node map
Definition graph_algorithms.hxx:117
detail::SelectIntegerType< 32, detail::UnsignedIntTypes >::type UInt32
32-bit unsigned int
Definition sized_int.hxx:183
void fillNodeMap(const G &g, A &a, const T &value)
fill a lemon node map
Definition graph_algorithms.hxx:136
bool allLessEqual(TinyVectorBase< V1, SIZE, D1, D2 > const &l, TinyVectorBase< V2, SIZE, D3, D4 > const &r)
pointwise less-equal
Definition tinyvector.hxx:1399
void makeRegionAdjacencyGraph(GRAPH_IN graphIn, GRAPH_IN_NODE_LABEL_MAP labels, AdjacencyListGraph &rag, typename AdjacencyListGraph::template EdgeMap< std::vector< typename GRAPH_IN::Edge > > &affiliatedEdges, const Int64 ignoreLabel=-1)
make a region adjacency graph from a graph and labels w.r.t. that graph
Definition graph_algorithms.hxx:165
void carvingSegmentation(const GRAPH &g, const EDGE_WEIGHTS &edgeWeights, const SEEDS &seeds, const typename LABELS::Value backgroundLabel, const typename EDGE_WEIGHTS::Value backgroundBias, const typename EDGE_WEIGHTS::Value noPriorBelow, LABELS &labels)
edge weighted watersheds Segmentataion
Definition graph_algorithms.hxx:892
size_t pathLength(const NODE source, const NODE target, const PREDECESSORS &predecessors)
get the length in node units of a path
Definition graph_algorithms.hxx:591
FFTWComplex< R >::NormType norm(const FFTWComplex< R > &a)
norm (= magnitude)
Definition fftw3.hxx:1037
void recursiveGraphSmoothing(const GRAPH &g, const NODE_FEATURES_IN &nodeFeaturesIn, const EDGE_INDICATOR &edgeIndicator, const float lambda, const float edgeThreshold, const float scale, size_t iterations, NODE_FEATURES_OUT &nodeFeaturesBuffer, NODE_FEATURES_OUT &nodeFeaturesOut)
smooth node features of a graph
Definition graph_algorithms.hxx:1124
void copyEdgeMap(const G &g, const A &a, B &b)
copy a lemon edge map
Definition graph_algorithms.hxx:127
detail::SelectIntegerType< 64, detail::SignedIntTypes >::type Int64
64-bit signed int
Definition sized_int.hxx:177
void edgeSort(const GRAPH &g, const WEIGHTS &weights, const COMPERATOR &comperator, std::vector< typename GRAPH::Edge > &sortedEdges)
get a vector of Edge descriptors
Definition graph_algorithms.hxx:98
void fillEdgeMap(const G &g, A &a, const T &value)
fill a lemon edge map
Definition graph_algorithms.hxx:145
void edgeWeightedWatershedsSegmentation(const GRAPH &g, const EDGE_WEIGHTS &edgeWeights, const SEEDS &seeds, LABELS &labels)
edge weighted watersheds Segmentataion
Definition graph_algorithms.hxx:871
void felzenszwalbSegmentation(const GRAPH &graph, const EDGE_WEIGHTS &edgeWeights, const NODE_SIZE &nodeSizes, float k, NODE_LABEL_MAP &nodeLabeling, const int nodeNumStopCond=-1)
edge weighted watersheds Segmentataion
Definition graph_algorithms.hxx:916
bool allLess(TinyVectorBase< V1, SIZE, D1, D2 > const &l, TinyVectorBase< V2, SIZE, D3, D4 > const &r)
pointwise less-than
Definition tinyvector.hxx:1375
MultiArray< 2, MultiArrayIndex > ragFindEdges(const RAGGRAPH &rag, const GRAPH &graph, const RAGEDGES &affiliatedEdges, MultiArrayView< N, T > labelsArray, const typename RAGGRAPH::Node &node)
Find indices of points on the edges.
Definition graph_algorithms.hxx:1268
void edgeWeightsFromInterpolatedImage(const GridGraph< N, DirectedTag > &g, const MultiArrayView< N, T > &interpolatedImage, EDGEMAP &edgeWeights, bool euclidean=false)
create edge weights from an interpolated image
Definition graph_algorithms.hxx:1399
void graphSmoothing(const GRAPH &g, const NODE_FEATURES_IN &nodeFeaturesIn, const EDGE_INDICATOR &edgeIndicator, const float lambda, const float edgeThreshold, const float scale, NODE_FEATURES_OUT &nodeFeaturesOut)
smooth node features of a graph
Definition graph_algorithms.hxx:1099
void edgeWeightsFromNodeWeights(const GridGraph< N, DirectedTag > &g, const NODEMAP &nodeWeights, EDGEMAP &edgeWeights, bool euclidean, FUNCTOR const &func)
create edge weights from node weights
Definition graph_algorithms.hxx:1341
void shortestPathAStar(const GRAPH &graph, const typename GRAPH::Node &source, const typename GRAPH::Node &target, const WEIGHTS &weights, PREDECESSORS &predecessors, DISTANCE &distance, const HEURSTIC &heuristic)
Astar Shortest path search.
Definition graph_algorithms.hxx:611
void initMultiArrayBorder(...)
Write values to the specified border values in the array.

© Ullrich Köthe (ullrich.koethe@iwr.uni-heidelberg.de)
Heidelberg Collaboratory for Image Processing, University of Heidelberg, Germany

html generated using doxygen and Python
vigra 1.11.2