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//                        Intel License Agreement
11//                For Open Source Computer Vision Library
12//
13// Copyright (C) 2000, Intel Corporation, all rights reserved.
14// Third party copyrights are property of their respective owners.
15//
16// Redistribution and use in source and binary forms, with or without modification,
17// are permitted provided that the following conditions are met:
18//
19//   * Redistribution's of source code must retain the above copyright notice,
20//     this list of conditions and the following disclaimer.
21//
22//   * Redistribution's in binary form must reproduce the above copyright notice,
23//     this list of conditions and the following disclaimer in the documentation
24//     and/or other materials provided with the distribution.
25//
26//   * The name of Intel Corporation may not be used to endorse or promote products
27//     derived from this software without specific prior written permission.
28//
29// This software is provided by the copyright holders and contributors "as is" and
30// any express or implied warranties, including, but not limited to, the implied
31// warranties of merchantability and fitness for a particular purpose are disclaimed.
32// In no event shall the Intel Corporation or contributors be liable for any direct,
33// indirect, incidental, special, exemplary, or consequential damages
34// (including, but not limited to, procurement of substitute goods or services;
35// loss of use, data, or profits; or business interruption) however caused
36// and on any theory of liability, whether in contract, strict liability,
37// or tort (including negligence or otherwise) arising in any way out of
38// the use of this software, even if advised of the possibility of such damage.
39//
40//M*/
41
42#include "test_precomp.hpp"
43
44using namespace cv;
45using namespace std;
46
47class CV_TemplMatchTest : public cvtest::ArrayTest
48{
49public:
50    CV_TemplMatchTest();
51
52protected:
53    int read_params( CvFileStorage* fs );
54    void get_test_array_types_and_sizes( int test_case_idx, vector<vector<Size> >& sizes, vector<vector<int> >& types );
55    void get_minmax_bounds( int i, int j, int type, Scalar& low, Scalar& high );
56    double get_success_error_level( int test_case_idx, int i, int j );
57    void run_func();
58    void prepare_to_validation( int );
59
60    int max_template_size;
61    int method;
62    bool test_cpp;
63};
64
65
66CV_TemplMatchTest::CV_TemplMatchTest()
67{
68    test_array[INPUT].push_back(NULL);
69    test_array[INPUT].push_back(NULL);
70    test_array[OUTPUT].push_back(NULL);
71    test_array[REF_OUTPUT].push_back(NULL);
72    element_wise_relative_error = false;
73    max_template_size = 100;
74    method = 0;
75    test_cpp = false;
76}
77
78
79int CV_TemplMatchTest::read_params( CvFileStorage* fs )
80{
81    int code = cvtest::ArrayTest::read_params( fs );
82    if( code < 0 )
83        return code;
84
85    max_template_size = cvReadInt( find_param( fs, "max_template_size" ), max_template_size );
86    max_template_size = cvtest::clipInt( max_template_size, 1, 100 );
87
88    return code;
89}
90
91
92void CV_TemplMatchTest::get_minmax_bounds( int i, int j, int type, Scalar& low, Scalar& high )
93{
94    cvtest::ArrayTest::get_minmax_bounds( i, j, type, low, high );
95    int depth = CV_MAT_DEPTH(type);
96    if( depth == CV_32F )
97    {
98        low = Scalar::all(-10.);
99        high = Scalar::all(10.);
100    }
101}
102
103
104void CV_TemplMatchTest::get_test_array_types_and_sizes( int test_case_idx,
105                                                vector<vector<Size> >& sizes, vector<vector<int> >& types )
106{
107    RNG& rng = ts->get_rng();
108    int depth = cvtest::randInt(rng) % 2, cn = cvtest::randInt(rng) & 1 ? 3 : 1;
109    cvtest::ArrayTest::get_test_array_types_and_sizes( test_case_idx, sizes, types );
110    depth = depth == 0 ? CV_8U : CV_32F;
111
112    types[INPUT][0] = types[INPUT][1] = CV_MAKETYPE(depth,cn);
113    types[OUTPUT][0] = types[REF_OUTPUT][0] = CV_32FC1;
114
115    sizes[INPUT][1].width = cvtest::randInt(rng)%MIN(sizes[INPUT][1].width,max_template_size) + 1;
116    sizes[INPUT][1].height = cvtest::randInt(rng)%MIN(sizes[INPUT][1].height,max_template_size) + 1;
117    sizes[OUTPUT][0].width = sizes[INPUT][0].width - sizes[INPUT][1].width + 1;
118    sizes[OUTPUT][0].height = sizes[INPUT][0].height - sizes[INPUT][1].height + 1;
119    sizes[REF_OUTPUT][0] = sizes[OUTPUT][0];
120
121    method = cvtest::randInt(rng)%6;
122    test_cpp = (cvtest::randInt(rng) & 256) == 0;
123}
124
125
126double CV_TemplMatchTest::get_success_error_level( int /*test_case_idx*/, int /*i*/, int /*j*/ )
127{
128    if( test_mat[INPUT][1].depth() == CV_8U ||
129        (method >= CV_TM_CCOEFF && test_mat[INPUT][1].cols*test_mat[INPUT][1].rows <= 2) )
130        return 1e-2;
131    else
132        return 1e-3;
133}
134
135
136void CV_TemplMatchTest::run_func()
137{
138    if(!test_cpp)
139        cvMatchTemplate( test_array[INPUT][0], test_array[INPUT][1], test_array[OUTPUT][0], method );
140    else
141    {
142        cv::Mat _out = cv::cvarrToMat(test_array[OUTPUT][0]);
143        cv::matchTemplate(cv::cvarrToMat(test_array[INPUT][0]), cv::cvarrToMat(test_array[INPUT][1]), _out, method);
144    }
145}
146
147
148static void cvTsMatchTemplate( const CvMat* img, const CvMat* templ, CvMat* result, int method )
149{
150    int i, j, k, l;
151    int depth = CV_MAT_DEPTH(img->type), cn = CV_MAT_CN(img->type);
152    int width_n = templ->cols*cn, height = templ->rows;
153    int a_step = img->step / CV_ELEM_SIZE(img->type & CV_MAT_DEPTH_MASK);
154    int b_step = templ->step / CV_ELEM_SIZE(templ->type & CV_MAT_DEPTH_MASK);
155    CvScalar b_mean, b_sdv;
156    double b_denom = 1., b_sum2 = 0;
157    int area = templ->rows*templ->cols;
158
159    cvAvgSdv(templ, &b_mean, &b_sdv);
160
161    for( i = 0; i < cn; i++ )
162        b_sum2 += (b_sdv.val[i]*b_sdv.val[i] + b_mean.val[i]*b_mean.val[i])*area;
163
164    if( b_sdv.val[0]*b_sdv.val[0] + b_sdv.val[1]*b_sdv.val[1] +
165        b_sdv.val[2]*b_sdv.val[2] + b_sdv.val[3]*b_sdv.val[3] < DBL_EPSILON &&
166        method == CV_TM_CCOEFF_NORMED )
167    {
168        cvSet( result, cvScalarAll(1.) );
169        return;
170    }
171
172    if( method & 1 )
173    {
174        b_denom = 0;
175        if( method != CV_TM_CCOEFF_NORMED )
176        {
177            b_denom = b_sum2;
178        }
179        else
180        {
181            for( i = 0; i < cn; i++ )
182                b_denom += b_sdv.val[i]*b_sdv.val[i]*area;
183        }
184        b_denom = sqrt(b_denom);
185        if( b_denom == 0 )
186            b_denom = 1.;
187    }
188
189    assert( CV_TM_SQDIFF <= method && method <= CV_TM_CCOEFF_NORMED );
190
191    for( i = 0; i < result->rows; i++ )
192    {
193        for( j = 0; j < result->cols; j++ )
194        {
195            CvScalar a_sum(0), a_sum2(0);
196            CvScalar ccorr(0);
197            double value = 0.;
198
199            if( depth == CV_8U )
200            {
201                const uchar* a = img->data.ptr + i*img->step + j*cn;
202                const uchar* b = templ->data.ptr;
203
204                if( cn == 1 || method < CV_TM_CCOEFF )
205                {
206                    for( k = 0; k < height; k++, a += a_step, b += b_step )
207                        for( l = 0; l < width_n; l++ )
208                        {
209                            ccorr.val[0] += a[l]*b[l];
210                            a_sum.val[0] += a[l];
211                            a_sum2.val[0] += a[l]*a[l];
212                        }
213                }
214                else
215                {
216                    for( k = 0; k < height; k++, a += a_step, b += b_step )
217                        for( l = 0; l < width_n; l += 3 )
218                        {
219                            ccorr.val[0] += a[l]*b[l];
220                            ccorr.val[1] += a[l+1]*b[l+1];
221                            ccorr.val[2] += a[l+2]*b[l+2];
222                            a_sum.val[0] += a[l];
223                            a_sum.val[1] += a[l+1];
224                            a_sum.val[2] += a[l+2];
225                            a_sum2.val[0] += a[l]*a[l];
226                            a_sum2.val[1] += a[l+1]*a[l+1];
227                            a_sum2.val[2] += a[l+2]*a[l+2];
228                        }
229                }
230            }
231            else
232            {
233                const float* a = (const float*)(img->data.ptr + i*img->step) + j*cn;
234                const float* b = (const float*)templ->data.ptr;
235
236                if( cn == 1 || method < CV_TM_CCOEFF )
237                {
238                    for( k = 0; k < height; k++, a += a_step, b += b_step )
239                        for( l = 0; l < width_n; l++ )
240                        {
241                            ccorr.val[0] += a[l]*b[l];
242                            a_sum.val[0] += a[l];
243                            a_sum2.val[0] += a[l]*a[l];
244                        }
245                }
246                else
247                {
248                    for( k = 0; k < height; k++, a += a_step, b += b_step )
249                        for( l = 0; l < width_n; l += 3 )
250                        {
251                            ccorr.val[0] += a[l]*b[l];
252                            ccorr.val[1] += a[l+1]*b[l+1];
253                            ccorr.val[2] += a[l+2]*b[l+2];
254                            a_sum.val[0] += a[l];
255                            a_sum.val[1] += a[l+1];
256                            a_sum.val[2] += a[l+2];
257                            a_sum2.val[0] += a[l]*a[l];
258                            a_sum2.val[1] += a[l+1]*a[l+1];
259                            a_sum2.val[2] += a[l+2]*a[l+2];
260                        }
261                }
262            }
263
264            switch( method )
265            {
266            case CV_TM_CCORR:
267            case CV_TM_CCORR_NORMED:
268                value = ccorr.val[0];
269                break;
270            case CV_TM_SQDIFF:
271            case CV_TM_SQDIFF_NORMED:
272                value = (a_sum2.val[0] + b_sum2 - 2*ccorr.val[0]);
273                break;
274            default:
275                value = (ccorr.val[0] - a_sum.val[0]*b_mean.val[0]+
276                         ccorr.val[1] - a_sum.val[1]*b_mean.val[1]+
277                         ccorr.val[2] - a_sum.val[2]*b_mean.val[2]);
278            }
279
280            if( method & 1 )
281            {
282                double denom;
283
284                // calc denominator
285                if( method != CV_TM_CCOEFF_NORMED )
286                {
287                    denom = a_sum2.val[0] + a_sum2.val[1] + a_sum2.val[2];
288                }
289                else
290                {
291                    denom = a_sum2.val[0] - (a_sum.val[0]*a_sum.val[0])/area;
292                    denom += a_sum2.val[1] - (a_sum.val[1]*a_sum.val[1])/area;
293                    denom += a_sum2.val[2] - (a_sum.val[2]*a_sum.val[2])/area;
294                }
295                denom = sqrt(MAX(denom,0))*b_denom;
296                if( fabs(value) < denom )
297                    value /= denom;
298                else if( fabs(value) < denom*1.125 )
299                    value = value > 0 ? 1 : -1;
300                else
301                    value = method != CV_TM_SQDIFF_NORMED ? 0 : 1;
302            }
303
304            ((float*)(result->data.ptr + result->step*i))[j] = (float)value;
305        }
306    }
307}
308
309
310void CV_TemplMatchTest::prepare_to_validation( int /*test_case_idx*/ )
311{
312    CvMat _input = test_mat[INPUT][0], _templ = test_mat[INPUT][1];
313    CvMat _output = test_mat[REF_OUTPUT][0];
314    cvTsMatchTemplate( &_input, &_templ, &_output, method );
315
316    //if( ts->get_current_test_info()->test_case_idx == 0 )
317    /*{
318        CvFileStorage* fs = cvOpenFileStorage( "_match_template.yml", 0, CV_STORAGE_WRITE );
319        cvWrite( fs, "image", &test_mat[INPUT][0] );
320        cvWrite( fs, "template", &test_mat[INPUT][1] );
321        cvWrite( fs, "ref", &test_mat[REF_OUTPUT][0] );
322        cvWrite( fs, "opencv", &test_mat[OUTPUT][0] );
323        cvWriteInt( fs, "method", method );
324        cvReleaseFileStorage( &fs );
325    }*/
326
327    if( method >= CV_TM_CCOEFF )
328    {
329        // avoid numerical stability problems in singular cases (when the results are near to 0)
330        const double delta = 10.;
331        test_mat[REF_OUTPUT][0] += Scalar::all(delta);
332        test_mat[OUTPUT][0] += Scalar::all(delta);
333    }
334}
335
336TEST(Imgproc_MatchTemplate, accuracy) { CV_TemplMatchTest test; test.safe_run(); }
337