• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*M///////////////////////////////////////////////////////////////////////////////////////
2 //
3 //  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
4 //
5 //  By downloading, copying, installing or using the software you agree to this license.
6 //  If you do not agree to this license, do not download, install,
7 //  copy or use the software.
8 //
9 //
10 //                           License Agreement
11 //                For Open Source Computer Vision Library
12 //
13 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
14 // Copyright (C) 2009, Willow Garage Inc., all rights reserved.
15 // Third party copyrights are property of their respective owners.
16 //
17 // Redistribution and use in source and binary forms, with or without modification,
18 // are permitted provided that the following conditions are met:
19 //
20 //   * Redistribution's of source code must retain the above copyright notice,
21 //     this list of conditions and the following disclaimer.
22 //
23 //   * Redistribution's in binary form must reproduce the above copyright notice,
24 //     this list of conditions and the following disclaimer in the documentation
25 //     and/or other materials provided with the distribution.
26 //
27 //   * The name of the copyright holders may not be used to endorse or promote products
28 //     derived from this software without specific prior written permission.
29 //
30 // This software is provided by the copyright holders and contributors "as is" and
31 // any express or implied warranties, including, but not limited to, the implied
32 // warranties of merchantability and fitness for a particular purpose are disclaimed.
33 // In no event shall the Intel Corporation or contributors be liable for any direct,
34 // indirect, incidental, special, exemplary, or consequential damages
35 // (including, but not limited to, procurement of substitute goods or services;
36 // loss of use, data, or profits; or business interruption) however caused
37 // and on any theory of liability, whether in contract, strict liability,
38 // or tort (including negligence or otherwise) arising in any way out of
39 // the use of this software, even if advised of the possibility of such damage.
40 //
41 //M*/
42 
43 #include "precomp.hpp"
44 
45 namespace cv
46 {
47 
48 class ThinPlateSplineShapeTransformerImpl : public ThinPlateSplineShapeTransformer
49 {
50 public:
51     /* Constructors */
ThinPlateSplineShapeTransformerImpl()52     ThinPlateSplineShapeTransformerImpl()
53     {
54         regularizationParameter=0;
55         name_ = "ShapeTransformer.TPS";
56         tpsComputed=false;
57     }
58 
ThinPlateSplineShapeTransformerImpl(double _regularizationParameter)59     ThinPlateSplineShapeTransformerImpl(double _regularizationParameter)
60     {
61         regularizationParameter=_regularizationParameter;
62         name_ = "ShapeTransformer.TPS";
63         tpsComputed=false;
64     }
65 
66     /* Destructor */
~ThinPlateSplineShapeTransformerImpl()67     ~ThinPlateSplineShapeTransformerImpl()
68     {
69     }
70 
71     //! the main operators
72     virtual void estimateTransformation(InputArray transformingShape, InputArray targetShape, std::vector<DMatch> &matches);
73     virtual float applyTransformation(InputArray inPts, OutputArray output=noArray());
74     virtual void warpImage(InputArray transformingImage, OutputArray output,
75                            int flags, int borderMode, const Scalar& borderValue) const;
76 
77     //! Setters/Getters
setRegularizationParameter(double _regularizationParameter)78     virtual void setRegularizationParameter(double _regularizationParameter) {regularizationParameter=_regularizationParameter;}
getRegularizationParameter() const79     virtual double getRegularizationParameter() const {return regularizationParameter;}
80 
81     //! write/read
write(FileStorage & fs) const82     virtual void write(FileStorage& fs) const
83     {
84         fs << "name" << name_
85            << "regularization" << regularizationParameter;
86     }
87 
read(const FileNode & fn)88     virtual void read(const FileNode& fn)
89     {
90         CV_Assert( (String)fn["name"] == name_ );
91         regularizationParameter = (int)fn["regularization"];
92     }
93 
94 private:
95     bool tpsComputed;
96     double regularizationParameter;
97     float transformCost;
98     Mat tpsParameters;
99     Mat shapeReference;
100 
101 protected:
102     String name_;
103 };
104 
distance(Point2f p,Point2f q)105 static float distance(Point2f p, Point2f q)
106 {
107     Point2f diff = p - q;
108     float norma = diff.x*diff.x + diff.y*diff.y;// - 2*diff.x*diff.y;
109     if (norma<0) norma=0;
110     //else norma = std::sqrt(norma);
111     norma = norma*std::log(norma+FLT_EPSILON);
112     return norma;
113 }
114 
_applyTransformation(const Mat & shapeRef,const Point2f point,const Mat & tpsParameters)115 static Point2f _applyTransformation(const Mat &shapeRef, const Point2f point, const Mat &tpsParameters)
116 {
117     Point2f out;
118     for (int i=0; i<2; i++)
119     {
120         float a1=tpsParameters.at<float>(tpsParameters.rows-3,i);
121         float ax=tpsParameters.at<float>(tpsParameters.rows-2,i);
122         float ay=tpsParameters.at<float>(tpsParameters.rows-1,i);
123 
124         float affine=a1+ax*point.x+ay*point.y;
125         float nonrigid=0;
126         for (int j=0; j<shapeRef.rows; j++)
127         {
128             nonrigid+=tpsParameters.at<float>(j,i)*
129                     distance(Point2f(shapeRef.at<float>(j,0),shapeRef.at<float>(j,1)),
130                             point);
131         }
132         if (i==0)
133         {
134             out.x=affine+nonrigid;
135         }
136         if (i==1)
137         {
138             out.y=affine+nonrigid;
139         }
140     }
141     return out;
142 }
143 
144 /* public methods */
warpImage(InputArray transformingImage,OutputArray output,int flags,int borderMode,const Scalar & borderValue) const145 void ThinPlateSplineShapeTransformerImpl::warpImage(InputArray transformingImage, OutputArray output,
146                                       int flags, int borderMode, const Scalar& borderValue) const
147 {
148     CV_Assert(tpsComputed==true);
149 
150     Mat theinput = transformingImage.getMat();
151     Mat mapX(theinput.rows, theinput.cols, CV_32FC1);
152     Mat mapY(theinput.rows, theinput.cols, CV_32FC1);
153 
154     for (int row = 0; row < theinput.rows; row++)
155     {
156         for (int col = 0; col < theinput.cols; col++)
157         {
158             Point2f pt = _applyTransformation(shapeReference, Point2f(float(col), float(row)), tpsParameters);
159             mapX.at<float>(row, col) = pt.x;
160             mapY.at<float>(row, col) = pt.y;
161         }
162     }
163     remap(transformingImage, output, mapX, mapY, flags, borderMode, borderValue);
164 }
165 
applyTransformation(InputArray inPts,OutputArray outPts)166 float ThinPlateSplineShapeTransformerImpl::applyTransformation(InputArray inPts, OutputArray outPts)
167 {
168     CV_Assert(tpsComputed);
169     Mat pts1 = inPts.getMat();
170     CV_Assert((pts1.channels()==2) && (pts1.cols>0));
171 
172     //Apply transformation in the complete set of points
173     // Ensambling output //
174     if (outPts.needed())
175     {
176         outPts.create(1,pts1.cols, CV_32FC2);
177         Mat outMat = outPts.getMat();
178         for (int i=0; i<pts1.cols; i++)
179         {
180             Point2f pt=pts1.at<Point2f>(0,i);
181             outMat.at<Point2f>(0,i)=_applyTransformation(shapeReference, pt, tpsParameters);
182         }
183     }
184 
185     return transformCost;
186 }
187 
estimateTransformation(InputArray _pts1,InputArray _pts2,std::vector<DMatch> & _matches)188 void ThinPlateSplineShapeTransformerImpl::estimateTransformation(InputArray _pts1, InputArray _pts2,
189                                                                std::vector<DMatch>& _matches )
190 {
191     Mat pts1 = _pts1.getMat();
192     Mat pts2 = _pts2.getMat();
193     CV_Assert((pts1.channels()==2) && (pts1.cols>0) && (pts2.channels()==2) && (pts2.cols>0));
194     CV_Assert(_matches.size()>1);
195 
196     if (pts1.type() != CV_32F)
197         pts1.convertTo(pts1, CV_32F);
198     if (pts2.type() != CV_32F)
199         pts2.convertTo(pts2, CV_32F);
200 
201     // Use only valid matchings //
202     std::vector<DMatch> matches;
203     for (size_t i=0; i<_matches.size(); i++)
204     {
205         if (_matches[i].queryIdx<pts1.cols &&
206             _matches[i].trainIdx<pts2.cols)
207         {
208             matches.push_back(_matches[i]);
209         }
210     }
211 
212     // Organizing the correspondent points in matrix style //
213     Mat shape1((int)matches.size(),2,CV_32F); // transforming shape
214     Mat shape2((int)matches.size(),2,CV_32F); // target shape
215     for (int i=0, end = (int)matches.size(); i<end; i++)
216     {
217         Point2f pt1=pts1.at<Point2f>(0,matches[i].queryIdx);
218         shape1.at<float>(i,0) = pt1.x;
219         shape1.at<float>(i,1) = pt1.y;
220 
221         Point2f pt2=pts2.at<Point2f>(0,matches[i].trainIdx);
222         shape2.at<float>(i,0) = pt2.x;
223         shape2.at<float>(i,1) = pt2.y;
224     }
225     shape1.copyTo(shapeReference);
226 
227     // Building the matrices for solving the L*(w|a)=(v|0) problem with L={[K|P];[P'|0]}
228 
229     //Building K and P (Neede to buil L)
230     Mat matK((int)matches.size(),(int)matches.size(),CV_32F);
231     Mat matP((int)matches.size(),3,CV_32F);
232     for (int i=0, end=(int)matches.size(); i<end; i++)
233     {
234         for (int j=0; j<end; j++)
235         {
236             if (i==j)
237             {
238                 matK.at<float>(i,j)=float(regularizationParameter);
239             }
240             else
241             {
242                 matK.at<float>(i,j) = distance(Point2f(shape1.at<float>(i,0),shape1.at<float>(i,1)),
243                                                Point2f(shape1.at<float>(j,0),shape1.at<float>(j,1)));
244             }
245         }
246         matP.at<float>(i,0) = 1;
247         matP.at<float>(i,1) = shape1.at<float>(i,0);
248         matP.at<float>(i,2) = shape1.at<float>(i,1);
249     }
250 
251     //Building L
252     Mat matL=Mat::zeros((int)matches.size()+3,(int)matches.size()+3,CV_32F);
253     Mat matLroi(matL, Rect(0,0,(int)matches.size(),(int)matches.size())); //roi for K
254     matK.copyTo(matLroi);
255     matLroi = Mat(matL,Rect((int)matches.size(),0,3,(int)matches.size())); //roi for P
256     matP.copyTo(matLroi);
257     Mat matPt;
258     transpose(matP,matPt);
259     matLroi = Mat(matL,Rect(0,(int)matches.size(),(int)matches.size(),3)); //roi for P'
260     matPt.copyTo(matLroi);
261 
262     //Building B (v|0)
263     Mat matB = Mat::zeros((int)matches.size()+3,2,CV_32F);
264     for (int i=0, end = (int)matches.size(); i<end; i++)
265     {
266         matB.at<float>(i,0) = shape2.at<float>(i,0); //x's
267         matB.at<float>(i,1) = shape2.at<float>(i,1); //y's
268     }
269 
270     //Obtaining transformation params (w|a)
271     solve(matL, matB, tpsParameters, DECOMP_LU);
272     //tpsParameters = matL.inv()*matB;
273 
274     //Setting transform Cost and Shape reference
275     Mat w(tpsParameters, Rect(0,0,2,tpsParameters.rows-3));
276     Mat Q=w.t()*matK*w;
277     transformCost=fabs(Q.at<float>(0,0)*Q.at<float>(1,1));//fabs(mean(Q.diag(0))[0]);//std::max(Q.at<float>(0,0),Q.at<float>(1,1));
278     tpsComputed=true;
279 }
280 
createThinPlateSplineShapeTransformer(double regularizationParameter)281 Ptr <ThinPlateSplineShapeTransformer> createThinPlateSplineShapeTransformer(double regularizationParameter)
282 {
283     return Ptr<ThinPlateSplineShapeTransformer>( new ThinPlateSplineShapeTransformerImpl(regularizationParameter) );
284 }
285 
286 } // cv
287