-
Notifications
You must be signed in to change notification settings - Fork 134
/
Copy pathvideo_stitcher.cpp
1392 lines (1206 loc) · 39.3 KB
/
video_stitcher.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#include "stdafx.h"
#include <iostream>
#include <fstream>
#include <string>
#include <time.h>
#include <Windows.h>
#include "composer_gpu.h"
#include "video_stitcher.h"
#include "apap.h"
#include "gutil.h"
MyVideoStitcher::MyVideoStitcher()
{
is_preview_ = true;
is_save_video_ = false;
start_frame_index_ = 0;
end_frame_index_ = -1;
is_try_gpu_ = false;
is_debug_ = false;
trim_type_ = MyVideoStitcher::TRIM_NO;
work_megapix_ = 1.0;//-1;//
seam_megapix_ = 0.2;//-1;//
is_prepared_ = false;
conf_thresh_ = 1.f;
features_type_ = "orb";//"surf";//
ba_cost_func_ = "ray";
ba_refine_mask_ = "xxxxx";
is_do_wave_correct_ = true;
wave_correct_ = detail::WAVE_CORRECT_HORIZ;
is_save_graph_ = false;
warp_type_ = "cylindrical";//"plane";//"apap";//"paniniA2B1";//"transverseMercator";//"spherical";//
expos_comp_type_ = ExposureCompensator::GAIN_BLOCKS;//ExposureCompensator::GAIN;//
match_conf_ = 0.3f;
seam_find_type_ = "gc_color";//"voronoi";//
blend_type_ = Blender::FEATHER;//Blender::MULTI_BAND;//Blender::NO;//
blend_strength_ = 5;
// 获取当前系统的核数
SYSTEM_INFO sys_info;
GetSystemInfo(&sys_info);
parallel_num_ = sys_info.dwNumberOfProcessors;
}
int MyVideoStitcher::stitch( vector<VideoCapture> &captures, string &writer_file_name )
{
int video_num = captures.size();
vector<Mat> src(video_num);
Mat frame, dst, show_dst;
// Debug用信息
bool is_save_input_frames = false;
bool is_save_output_frames = true;
double fps = captures[0].get(CV_CAP_PROP_FPS);
// skip some frames
for(int j = 0; j < video_num; j++)
for(int i = 0; i < start_frame_index_; i++)
captures[j].read(frame);
// 第一帧，做一些初始化，并且确定结果视频的分辨率
for(int j = 0; j < video_num; j++)
{
if( !captures[j].read(frame))
return -1;
frame.copyTo(src[j]);
if(is_debug_)
{
char img_save_name[100];
sprintf(img_save_name, "/%d.jpg", j+1);
imwrite(debug_dir_path_ + img_save_name, src[j]);
}
}
long prepare_start_clock = clock();
int prepare_status = Prepare(src);
// 先用ORB特征测试，错误的话再使用SURF，仍然错误则报错，输入视频不符合条件
if( prepare_status == STITCH_CONFIG_ERROR )
{
cout << "video stitch config error!" << endl;
return -1;
}
if( prepare_status != STITCH_SUCCESS)
{
features_type_ = "surf";
cout << "video stitch first try failed, second try ... " << endl;
if( Prepare(src) != STITCH_SUCCESS)
{
cout << "videos input are invalid. Initialization failed." << endl;
return -1;
}
}
long prepare_end_clock = clock();
cout << "\tprepare time: " << prepare_end_clock - prepare_start_clock << "ms" << endl;
StitchFrame(src, dst);
if(is_debug_) //保存第一帧拼接结果和mask
{
imwrite(debug_dir_path_ + "/res.jpg", dst);
vector<Mat> img_masks(video_num);
for(int i = 0; i < video_num; i++)
{
img_masks[i].create(src[i].rows, src[i].cols, CV_8UC3);
img_masks[i].setTo(Scalar::all(255));
}
Mat dst_mask;
StitchFrame(img_masks, dst_mask);
imwrite(debug_dir_path_ + "/mask.jpg", dst_mask);
}
// 创建结果视频
VideoWriter writer;
if(is_save_video_)
{
writer.open(writer_file_name, CV_FOURCC('D', 'I','V', '3'), 20, Size(dst.cols, dst.rows));
writer.write(dst);
}
// 开始拼接
double stitch_time = 0;
FrameInfo frame_info;
frame_info.src.resize(video_num);
int frameidx = 1;
cout << "Stitching..." << endl;
string window_name = "视频拼接";
if(is_preview_)
namedWindow(window_name);
double show_scale = 1.0, scale_interval = 0.03;
int frame_show_interval = cvFloor(1000 / fps);
int failed_frame_count = 0;
char log_string[1000];
char log_file_name[200];
SYSTEMTIME sys_time = {0};
GetLocalTime(&sys_time);
sprintf(log_file_name, "%d%02d%02d-%02d%02d%02d.log",
sys_time.wYear, sys_time.wMonth, sys_time.wDay, sys_time.wHour, sys_time.wMinute, sys_time.wSecond);
ofstream log_file;
if(is_debug_)
log_file.open(debug_dir_path_ + log_file_name);
long long startTime = clock();
while(true)
{
long frame_time = 0;
// 采集
long cap_start_clock = clock();
int j;
for(j = 0; j < video_num; j++)
{
if( !captures[j].read(frame))
break;
frame.copyTo(frame_info.src[j]);
}
frame_info.frame_idx = frameidx;
frameidx++;
if(j != video_num || (end_frame_index_ >= 0 && frameidx >= end_frame_index_)) //有一个视频源结束，则停止拼接
break;
// 拼接
long stitch_start_clock = clock();
frame_info.stitch_status = StitchFrame(frame_info.src, frame_info.dst);
long stitch_clock = clock();
sprintf(log_string, "\tframe %d: stitch(%dms), capture(%dms)",
frame_info.frame_idx, stitch_clock - stitch_start_clock, stitch_start_clock - cap_start_clock);
printf("%s", log_string);
if(is_debug_)
log_file << log_string << endl;
stitch_time += stitch_clock - stitch_start_clock;
frame_time += stitch_clock - cap_start_clock;
// 拼接失败
if(frame_info.stitch_status != 0)
{
cout << "failed\n";
if(is_debug_)
log_file << "failed" << endl;
failed_frame_count++;
break;
}
// 保存视频
if(is_save_video_)
{
cout << ", write(";
if(is_save_output_frames)
{
char img_save_name[100];
sprintf(img_save_name, "/images/%d.jpg", frame_info.frame_idx);
imwrite(debug_dir_path_ + img_save_name, frame_info.dst);
}
long write_start_clock = clock();
writer.write(frame_info.dst);
long write_clock = clock();
cout << write_clock - write_start_clock << "ms)";
frame_time += write_clock - write_start_clock;
}
cout << endl;
// 显示---
if(is_preview_)
{
int key = waitKey(std::max(1, (int)(frame_show_interval - frame_time)));
if(key == 27) // ESC
break;
else if(key == 61 || key == 43) // +
show_scale += scale_interval;
else if(key == 45) // -
if(show_scale >= scale_interval)
show_scale -= scale_interval;
resize(frame_info.dst, show_dst, Size(show_scale * dst.cols, show_scale * dst.rows));
imshow(window_name, show_dst);
}
}
long long endTime = clock();
cout << "test " << endTime - startTime << endl;
cout << "\nStitch over" << endl;
cout << failed_frame_count << " frames failed." << endl;
cout << "\tfull view angle is " << cvRound(view_angle_) << "°" << endl;
if(is_debug_)
log_file << "\tfull view angle is " << cvRound(view_angle_) << "°" << endl;
writer.release();
cout << "\ton average: stitch time = " << stitch_time / (frameidx-1) << "ms" << endl;
cout << "\tcenter: (" << -dst_roi_.x << ", " << -dst_roi_.y << ")" << endl;
if(is_debug_)
{
log_file << "\ton average: stitch time = " << stitch_time / (frameidx-1) << "ms" << endl;
log_file << "\tcenter: (" << -dst_roi_.x << ", " << -dst_roi_.y << ")" << endl;
log_file.close();
}
return 0;
}
void MyVideoStitcher::InitMembers(int num_images)
{
}
/*
* 初始图像可能分辨率很高，先做一步降采样，可以提高时间效率
*/
void MyVideoStitcher::SetScales(vector<Mat> &src)
{
if(work_megapix_ < 0)
work_scale_ = 1.0;
else
work_scale_ = min(1.0, sqrt(work_megapix_ * 1e6 / src[0].size().area()));
if(seam_megapix_ < 0)
seam_scale_ = 1.0;
else
seam_scale_ = min(1.0, sqrt(seam_megapix_ * 1e6 / src[0].size().area()));
}
/*
* 特征提取，支持SURF和ORB
*/
int MyVideoStitcher::FindFeatures(vector<Mat> &src, vector<ImageFeatures> &features)
{
Ptr<FeaturesFinder> finder;
if (features_type_ == "surf")
{
#ifdef HAVE_OPENCV_GPU
if (is_try_gpu_ && gpu::getCudaEnabledDeviceCount() > 0)
finder = new SurfFeaturesFinderGpu();
else
#endif
finder = new SurfFeaturesFinder();
}
else if (features_type_ == "orb")
{
finder = new OrbFeaturesFinder();//Size(3,1), 1500, 1.3, 5);
}
else
{
cout << "Unknown 2D features type: '" << features_type_ << "'.\n";
return STITCH_CONFIG_ERROR;
}
int num_images = static_cast<int>(src.size());
Mat full_img, img;
for (int i = 0; i < num_images; ++i)
{
full_img = src[i].clone();//
if (work_megapix_ < 0)
img = full_img;
else
resize(full_img, img, Size(), work_scale_, work_scale_);
(*finder)(img, features[i]);
//LOGLN("Features in image #" << i+1 << "("<<img.size()<< "): " << features[i].keypoints.size());
features[i].img_idx = i;
}
finder->collectGarbage();
full_img.release();
img.release();
return STITCH_SUCCESS;
}
/*
* 特征匹配，然后去除噪声图片。本代码实现时，一旦出现噪声图片，就终止算法
* 返回值：
* 0 —— 正常
* -2 —— 存在噪声图片
*/
int MyVideoStitcher::MatchImages(vector<ImageFeatures> &features, vector<MatchesInfo> &pairwise_matches)
{
int total_num_images = static_cast<int>(features.size());
BestOf2NearestMatcher matcher(is_try_gpu_, match_conf_);
matcher(features, pairwise_matches);
matcher.collectGarbage();
// 去除噪声图像
vector<int> indices = leaveBiggestComponent(features, pairwise_matches, conf_thresh_);
// 一旦出现噪声图片，就终止算法
int num_images = static_cast<int>(indices.size());
if (num_images != total_num_images)
{
LOGLN(total_num_images - num_images << " videos are invaild");
return STITCH_NOISE;
}
return STITCH_SUCCESS;
}
/*
* 摄像机标定
*/
int MyVideoStitcher::CalibrateCameras(vector<ImageFeatures> &features, vector<MatchesInfo> &pairwise_matches, vector<CameraParams> &cameras)
{
HomographyBasedEstimator estimator;
Ptr<detail::BundleAdjusterBase> adjuster;
Mat_<uchar> refine_mask;
vector<double> focals;
estimator(features, pairwise_matches, cameras);
for (size_t i = 0; i < cameras.size(); ++i)
{
Mat R;
cameras[i].R.convertTo(R, CV_32F);
cameras[i].R = R;
//LOGLN("Initial intrinsics #" << i << ":\n" << cameras[i].K());
}
if (ba_cost_func_ == "reproj") adjuster = new detail::BundleAdjusterReproj();
else if (ba_cost_func_ == "ray") adjuster = new detail::BundleAdjusterRay();
else
{
cout << "Unknown bundle adjustment cost function: '" << ba_cost_func_ << "'.\n";
return STITCH_CONFIG_ERROR;
}
adjuster->setConfThresh(conf_thresh_);
refine_mask = Mat::zeros(3, 3, CV_8U);
if (ba_refine_mask_[0] == 'x') refine_mask(0,0) = 1;
if (ba_refine_mask_[1] == 'x') refine_mask(0,1) = 1;
if (ba_refine_mask_[2] == 'x') refine_mask(0,2) = 1;
if (ba_refine_mask_[3] == 'x') refine_mask(1,1) = 1;
if (ba_refine_mask_[4] == 'x') refine_mask(1,2) = 1;
adjuster->setRefinementMask(refine_mask);
(*adjuster)(features, pairwise_matches, cameras);
// Find median focal length
for (size_t i = 0; i < cameras.size(); ++i)
{
focals.push_back(cameras[i].focal);
//LOGLN("Camera #" << i+1 << ":\n" << cameras[i].t << cameras[i].R);
}
sort(focals.begin(), focals.end());
if (focals.size() % 2 == 1)
median_focal_len_ = static_cast<float>(focals[focals.size() / 2]);
else
median_focal_len_ = static_cast<float>(focals[focals.size() / 2 - 1] + focals[focals.size() / 2]) * 0.5f;
if (is_do_wave_correct_)
{
vector<Mat> rmats;
for (size_t i = 0; i < cameras.size(); ++i)
rmats.push_back(cameras[i].R);
waveCorrect(rmats, wave_correct_);
for (size_t i = 0; i < cameras.size(); ++i)
cameras[i].R = rmats[i];
}
if(is_debug_)
this->saveCameraParam(debug_dir_path_ + "/camera_param.dat");
return STITCH_SUCCESS;
}
/*
* 计算水平视角，用于判断是否适用于平面投影
*/
double MyVideoStitcher::GetViewAngle(vector<Mat> &src, vector<CameraParams> &cameras)
{
Ptr<WarperCreator> warper_creator = new cv::CylindricalWarper();
Ptr<RotationWarper> warper = warper_creator->create(median_focal_len_);
int num_images = static_cast<int>(src.size());
vector<Point> corners;
vector<Size> sizes;
for (int i = 0; i < num_images; ++i)
{
Mat_<float> K;
cameras[i].K().convertTo(K, CV_32F);
Rect roi = warper->warpRoi(Size(src[i].cols * work_scale_, src[i].rows * work_scale_), K, cameras[i].R);
corners.push_back(roi.tl());
sizes.push_back(roi.size());
}
Rect result_roi = resultRoi(corners, sizes);
double view_angle = result_roi.width * 180.0 / (median_focal_len_ * CV_PI);
return view_angle;
}
/*
* 计算接缝之前，需要先把原始图像和mask按照相机参数投影
*/
int MyVideoStitcher::WarpForSeam(vector<Mat> &src, vector<CameraParams> &cameras, vector<Mat> &masks_warped, vector<Mat> &images_warped)
{
// Warp images and their masks
#ifdef HAVE_OPENCV_GPU
if (is_try_gpu_ && gpu::getCudaEnabledDeviceCount() > 0)
{
if (warp_type_ == "plane") warper_creator_ = new cv::PlaneWarperGpu();
else if (warp_type_ == "cylindrical") warper_creator_ = new cv::CylindricalWarperGpu();
else if (warp_type_ == "spherical") warper_creator_ = new cv::SphericalWarperGpu();
}
else
#endif
{
if (warp_type_ == "plane") warper_creator_ = new cv::PlaneWarper();
else if (warp_type_ == "cylindrical") warper_creator_ = new cv::CylindricalWarper();
else if (warp_type_ == "spherical") warper_creator_ = new cv::SphericalWarper();
else if (warp_type_ == "fisheye") warper_creator_ = new cv::FisheyeWarper();
else if (warp_type_ == "stereographic") warper_creator_ = new cv::StereographicWarper();
else if (warp_type_ == "compressedPlaneA2B1") warper_creator_ = new cv::CompressedRectilinearWarper(2, 1);
else if (warp_type_ == "compressedPlaneA1.5B1") warper_creator_ = new cv::CompressedRectilinearWarper(1.5, 1);
else if (warp_type_ == "compressedPlanePortraitA2B1") warper_creator_ = new cv::CompressedRectilinearPortraitWarper(2, 1);
else if (warp_type_ == "compressedPlanePortraitA1.5B1") warper_creator_ = new cv::CompressedRectilinearPortraitWarper(1.5, 1);
else if (warp_type_ == "paniniA2B1") warper_creator_ = new cv::PaniniWarper(2, 1);
else if (warp_type_ == "paniniA1.5B1") warper_creator_ = new cv::PaniniWarper(1.5, 1);
else if (warp_type_ == "paniniPortraitA2B1") warper_creator_ = new cv::PaniniPortraitWarper(2, 1);
else if (warp_type_ == "paniniPortraitA1.5B1") warper_creator_ = new cv::PaniniPortraitWarper(1.5, 1);
else if (warp_type_ == "mercator") warper_creator_ = new cv::MercatorWarper();
else if (warp_type_ == "transverseMercator") warper_creator_ = new cv::TransverseMercatorWarper();
}
if (warper_creator_.empty())
{
cout << "Can't create the following warper '" << warp_type_ << "'\n";
return STITCH_CONFIG_ERROR;
}
float warp_scale = static_cast<float>(median_focal_len_ * seam_scale_ / work_scale_);
Ptr<RotationWarper> warper = warper_creator_->create(warp_scale);
int full_pano_width = cvFloor(warp_scale * 2 * CV_PI);
int num_images = static_cast<int>(src.size());
Mat img, mask;
for (int i = 0; i < num_images; ++i)
{
Mat_<float> K;
cameras[i].K().convertTo(K, CV_32F);
float swa = (float)seam_scale_ / work_scale_;
K(0,0) *= swa; K(0,2) *= swa;
K(1,1) *= swa; K(1,2) *= swa;
if (seam_megapix_ < 0)
img = src[i].clone();
else
resize(src[i], img, Size(), seam_scale_, seam_scale_);
mask.create(img.size(), CV_8U);
mask.setTo(Scalar::all(255));
Mat tmp_mask_warped, tmp_img_warped;
Point tmp_corner;
Size tmp_size;
warper->warp(mask, K, cameras[i].R, INTER_NEAREST, BORDER_CONSTANT, tmp_mask_warped);
// 考虑360度拼接的特殊情况
tmp_corner = warper->warp(img, K, cameras[i].R, INTER_LINEAR, BORDER_REFLECT, tmp_img_warped);
//cout << "warped width = " << tmp_mask_warped.cols << ", pano width = " << full_pano_width << endl;
if(abs(tmp_mask_warped.cols - full_pano_width) <= 10)
{
int x1, x2;
FindWidestInpaintRange(tmp_mask_warped, x1, x2);
Mat mask1, mask2, img1, img2;
Rect rect1(0, 0, x1, tmp_mask_warped.rows), rect2(x2+1, 0, tmp_mask_warped.cols-1-x2, tmp_mask_warped.rows);
tmp_mask_warped(rect1).copyTo(mask1);
tmp_mask_warped(rect2).copyTo(mask2);
masks_warped.push_back(mask1);
masks_warped.push_back(mask2);
tmp_img_warped(rect1).copyTo(img1);
tmp_img_warped(rect2).copyTo(img2);
images_warped.push_back(img1);
images_warped.push_back(img2);
corners_.push_back(tmp_corner);
corners_.push_back(tmp_corner + rect2.tl());
sizes_.push_back(rect1.size());
sizes_.push_back(rect2.size());
}
else
{
masks_warped.push_back(tmp_mask_warped);
corners_.push_back(tmp_corner);
images_warped.push_back(tmp_img_warped);
sizes_.push_back(tmp_img_warped.size());
}
}
return STITCH_SUCCESS;
}
/*
* 解决360°拼接问题。对于横跨360°接缝的图片，找到最宽的inpaint区域[x1, x2]
*/
int MyVideoStitcher::FindWidestInpaintRange(Mat mask, int &x1, int &x2)
{
vector<int> sum_row(mask.cols);
uchar *mask_ptr = mask.ptr<uchar>(0);
for(int x = 0; x < mask.cols; x++)
sum_row[x] = 0;
for(int x = 0; x < mask.cols; x++)
for(int y = 0; y < mask.rows; y++)
if(mask_ptr[y * mask.cols + x] != 0)
sum_row[x] = 1;
int cur_x1, cur_x2, max_range = 0;
for(int x = 1; x < mask.cols; x++) // 最左边肯定是1
{
if(sum_row[x - 1] == 1 && sum_row[x] == 0)
cur_x1 = x;
else if(sum_row[x - 1] == 0 && sum_row[x] == 1)
{
cur_x2 = x - 1;
if(cur_x2 - cur_x1 > max_range)
{
x1 = cur_x1;
x2 = cur_x2;
}
}
}
return 0;
}
/*
* 计算接缝
*/
int MyVideoStitcher::FindSeam(vector<Mat> &images_warped, vector<Mat> &masks_warped)
{
int num_images = static_cast<int>(images_warped.size());
vector<Mat> images_warped_f(num_images);
for (int i = 0; i < num_images; ++i)
images_warped[i].convertTo(images_warped_f[i], CV_32F);
Ptr<SeamFinder> seam_finder;
if (seam_find_type_ == "no")
seam_finder = new detail::NoSeamFinder();
else if (seam_find_type_ == "voronoi")
seam_finder = new detail::VoronoiSeamFinder();
else if (seam_find_type_ == "gc_color")
{
#ifdef HAVE_OPENCV_GPU
if (is_try_gpu_ && gpu::getCudaEnabledDeviceCount() > 0)
seam_finder = new detail::GraphCutSeamFinderGpu(GraphCutSeamFinderBase::COST_COLOR);
else
#endif
seam_finder = new detail::GraphCutSeamFinder(GraphCutSeamFinderBase::COST_COLOR);
}
else if (seam_find_type_ == "gc_colorgrad")
{
#ifdef HAVE_OPENCV_GPU
if (is_try_gpu_ && gpu::getCudaEnabledDeviceCount() > 0)
seam_finder = new detail::GraphCutSeamFinderGpu(GraphCutSeamFinderBase::COST_COLOR_GRAD);
else
#endif
seam_finder = new detail::GraphCutSeamFinder(GraphCutSeamFinderBase::COST_COLOR_GRAD);
}
else if (seam_find_type_ == "dp_color")
seam_finder = new detail::DpSeamFinder(DpSeamFinder::COLOR);
else if (seam_find_type_ == "dp_colorgrad")
seam_finder = new detail::DpSeamFinder(DpSeamFinder::COLOR_GRAD);
if (seam_finder.empty())
{
cout << "Can't create the following seam finder '" << seam_find_type_ << "'\n";
return STITCH_CONFIG_ERROR;
}
seam_finder->find(images_warped_f, corners_, masks_warped);
images_warped_f.clear();
return STITCH_SUCCESS;
}
/*
* 恢复原始图像大小
*/
int MyVideoStitcher::Rescale(vector<Mat> &src, vector<CameraParams> &cameras, vector<Mat> &seam_masks)
{
median_focal_len_ = median_focal_len_ / work_scale_;
Ptr<RotationWarper> warper = warper_creator_->create(median_focal_len_);
int full_pano_width = cvFloor(median_focal_len_ * 2 * CV_PI);
//cout << "median focal length: " << median_focal_len_ << endl;
// Update corners and sizes
int num_images = static_cast<int>(src.size());
Mat tmp_mask, tmp_dilated_mask, tmp_seam_mask;
corners_.clear();
sizes_.clear();
for (int src_idx = 0, seam_idx = 0; src_idx < num_images; ++src_idx)
{
// Update intrinsics
cameras[src_idx].focal /= work_scale_;
cameras[src_idx].ppx /= work_scale_;
cameras[src_idx].ppy /= work_scale_;
Mat K;
cameras[src_idx].K().convertTo(K, CV_32F);
// 计算最终image warp的坐标映射矩阵
Mat tmp_xmap, tmp_ymap;
warper->buildMaps(src[src_idx].size(), K, cameras[src_idx].R, tmp_xmap, tmp_ymap);
// Warp the current image mask
Mat tmp_mask_warped, tmp_final_blend_mask;
tmp_mask.create(src[src_idx].size(), CV_8U);
tmp_mask.setTo(Scalar::all(255));
Point tmp_corner = warper->warp(tmp_mask, K, cameras[src_idx].R, INTER_NEAREST, BORDER_CONSTANT, tmp_mask_warped);
// 考虑360度拼接的特殊情况
if(abs(tmp_mask_warped.cols - full_pano_width) <= 10)
{
int x1, x2;
FindWidestInpaintRange(tmp_mask_warped, x1, x2);
Mat warped_mask[2], blend_mask[2], xmap[2], ymap[2];
Rect rect[2];
rect[0] = Rect(0, 0, x1, tmp_mask_warped.rows);
rect[1] = Rect(x2+1, 0, tmp_mask_warped.cols-1-x2, tmp_mask_warped.rows);
for(int j = 0; j < 2; j++)
{
tmp_mask_warped(rect[j]).copyTo(warped_mask[j]);
final_warped_masks_.push_back(warped_mask[j]);
tmp_xmap(rect[j]).copyTo(xmap[j]);
xmaps_.push_back(xmap[j]);
tmp_ymap(rect[j]).copyTo(ymap[j]);
ymaps_.push_back(ymap[j]);
// 计算总的mask = warp_mask & seam_mask
dilate(seam_masks[seam_idx], tmp_dilated_mask, Mat()); //膨胀
resize(tmp_dilated_mask, tmp_seam_mask, rect[j].size());
final_blend_masks_.push_back(warped_mask[j] & tmp_seam_mask);
corners_.push_back(tmp_corner + rect[j].tl());
sizes_.push_back(rect[j].size());
src_indices_.push_back(src_idx);
seam_idx++;
}
}
else
{
xmaps_.push_back(tmp_xmap);
ymaps_.push_back(tmp_ymap);
final_warped_masks_.push_back(tmp_mask_warped);
corners_.push_back(tmp_corner);
Size sz = tmp_mask_warped.size();
sizes_.push_back(sz);
// 计算总的mask = warp_mask & seam_mask
dilate(seam_masks[seam_idx], tmp_dilated_mask, Mat()); //膨胀
resize(tmp_dilated_mask, tmp_seam_mask, sz);
final_blend_masks_.push_back(tmp_mask_warped & tmp_seam_mask);
src_indices_.push_back(src_idx);
seam_idx ++;
}
}
dst_roi_ = resultRoi(corners_, sizes_);
int parts_num = sizes_.size();
final_warped_images_.resize(parts_num);
for(int j = 0; j < parts_num; j++)
final_warped_images_[j].create(sizes_[j], src[src_indices_[j]].type());
tmp_dilated_mask.release();
tmp_seam_mask.release();
tmp_mask.release();
return STITCH_SUCCESS;
}
/*
* 拼接结果可能是不规则形状，裁剪成方形
*/
int MyVideoStitcher::TrimRect(Rect rect)
{
// 计算每幅图像的rect，并修改xmap和ymap
int top = rect.y;
int left = rect.x;
int bottom = rect.y + rect.height - 1;
int right = rect.x + rect.width - 1;
int num_images = xmaps_.size();
for(int i = 0; i < num_images; i++)
{
int top_i, bottom_i, left_i, right_i;
top_i = max(dst_roi_.y + top, corners_[i].y);
left_i = max(dst_roi_.x + left, corners_[i].x);
bottom_i = min(corners_[i].y + sizes_[i].height - 1, dst_roi_.y + bottom);
right_i = min(corners_[i].x + sizes_[i].width - 1, dst_roi_.x + right);
sizes_[i].height = bottom_i - top_i + 1;
sizes_[i].width = right_i - left_i + 1;
Rect map_rect(left_i - corners_[i].x, top_i - corners_[i].y,
sizes_[i].width, sizes_[i].height);
Mat tmp_map = xmaps_[i].clone();
tmp_map(map_rect).copyTo(xmaps_[i]);
tmp_map = ymaps_[i].clone();
tmp_map(map_rect).copyTo(ymaps_[i]);
Mat tmp_img = final_blend_masks_[i].clone();
tmp_img(map_rect).copyTo(final_blend_masks_[i]);
corners_[i].x = left_i;
corners_[i].y = top_i;
}
dst_roi_.x += left;
dst_roi_.y += top;
dst_roi_.width = right - left + 1;
dst_roi_.height = bottom - top + 1;
return STITCH_SUCCESS;
}
/*
* 如果是平面投影的话，可以自动去除未填充区域
*/
int MyVideoStitcher::TrimInpaint(vector<Mat> &src)
{
int num_images = static_cast<int>(src.size());
// 先计算最终图像的mask
dst_roi_ = resultRoi(corners_, sizes_);
Mat dst = Mat::zeros(dst_roi_.height, dst_roi_.width, CV_8UC1);
for(int i = 0; i < num_images; i++)
{
int dx = corners_[i].x - dst_roi_.x;
int dy = corners_[i].y - dst_roi_.y;
int img_rows = sizes_[i].height;
int img_cols = sizes_[i].width;
for(int y = 0; y < img_rows; y++)
{
uchar *mask_row_ptr = final_warped_masks_[i].ptr<uchar>(y);
uchar *dst_row_ptr = dst.ptr<uchar>(dy + y);
for(int x = 0; x < img_cols; x++)
dst_row_ptr[dx + x] += mask_row_ptr[x];
}
}
int x, y;
// top
for(y = 0; y < dst_roi_.height; y++)
{
uchar *dst_row_ptr = dst.ptr<uchar>(y);
if(!(this->IsRowCrossInpaint(dst_row_ptr, dst_roi_.width)))
break;
}
int top = y;
// bottom
for(y = dst_roi_.height - 1; y >= 0 ; y--)
{
uchar *dst_row_ptr = dst.ptr<uchar>(y);
if(!(this->IsRowCrossInpaint(dst_row_ptr, dst_roi_.width)))
break;
}
int bottom = y;
// left
uchar *dst_ptr_00 = dst.ptr<uchar>(0);
for(x = 0; x < dst_roi_.width; x++)
{
for(y = top; y < bottom; y++)
if(dst_ptr_00[y * (dst_roi_.width) + x] == 0)
break;
if(y == bottom)
break;
}
int left = x;
// right
for(x = dst_roi_.width-1; x >= 0 ; x--)
{
for(y = top; y < bottom; y++)
if(dst_ptr_00[y * (dst_roi_.width) + x] == 0)
break;
if(y == bottom)
break;
}
int right = x;
// 计算每幅图像的rect，并修改xmap和ymap
for(int i = 0; i < num_images; i++)
{
int top_i, bottom_i, left_i, right_i;
top_i = max(dst_roi_.y + top, corners_[i].y);
left_i = max(dst_roi_.x + left, corners_[i].x);
bottom_i = min(corners_[i].y + sizes_[i].height - 1, dst_roi_.y + bottom);
right_i = min(corners_[i].x + sizes_[i].width - 1, dst_roi_.x + right);
sizes_[i].height = bottom_i - top_i + 1;
sizes_[i].width = right_i - left_i + 1;
Rect rect(left_i - corners_[i].x, top_i - corners_[i].y,
sizes_[i].width, sizes_[i].height);
Mat tmp_map = xmaps_[i].clone();
tmp_map(rect).copyTo(xmaps_[i]);
tmp_map = ymaps_[i].clone();
tmp_map(rect).copyTo(ymaps_[i]);
Mat tmp_img = final_blend_masks_[i].clone();
tmp_img(rect).copyTo(final_blend_masks_[i]);
corners_[i].x = left_i;
corners_[i].y = top_i;
}
dst_roi_.x += left;
dst_roi_.y += top;
dst_roi_.width = right - left + 1;
dst_roi_.height = bottom - top + 1;
return 0;
}
/*
* 判断一行中是否有未填充像素
*/
bool MyVideoStitcher::IsRowCrossInpaint(uchar *row, int width)
{
bool is_have_entered_inpaint = false;
int count0 = 0;
for(int x = 1; x < width; x++)
{
if(row[x] == 0)
count0++;
if(row[x-1] != 0 && row[x] == 0)
is_have_entered_inpaint = true;
if((row[x-1] == 0 && row[x] != 0) && is_have_entered_inpaint)
return true;
}
if(count0 >= (width/2))
return true;
return false;
}
int MyVideoStitcher::Prepare(vector<Mat> &src)
{
cv::setBreakOnError(true);
int num_images = static_cast<int>(src.size());
if (num_images < 2)
{
LOGLN("Need more images");
return STITCH_CONFIG_ERROR;
}
int cudaStatus = testGPU();
if( is_try_gpu_ && cudaStatus != 0 )
{
LOGLN("GPU acceleration failed! Error code: " << cudaStatus << " Please ensure that you have a CUDA-capable GPU installed!");
LOGLN("Stitching with CPU next ...");
return STITCH_CONFIG_ERROR;
}
int flag;
if(warp_type_ == "apap")
flag = PrepareAPAP(src);
else
flag = PrepareClassical(src);
if(flag == 0 && is_try_gpu_)
{
flag = initGPU(num_images);
if(flag != 0)
return flag;
C2GInitData *c2g_data = new C2GInitData[num_images];
for(int i = 0; i < num_images; i++)
{
c2g_data[i].xmap = xmaps_[i].ptr<float>(0);
c2g_data[i].ymap = ymaps_[i].ptr<float>(0);
c2g_data[i].ec_weight = ec_weight_maps_[i].ptr<float>(0);
c2g_data[i].blend_weight = blend_weight_maps_[i].ptr<float>(0);
c2g_data[i].total_weight = total_weight_maps_[i].ptr<float>(0);
c2g_data[i].height = src[i].rows;
c2g_data[i].width = src[i].cols;
c2g_data[i].warped_height = xmaps_[i].rows;
c2g_data[i].warped_width = xmaps_[i].cols;
c2g_data[i].corner_x = corners_[i].x - dst_roi_.x;
c2g_data[i].corner_y = corners_[i].y - dst_roi_.y;
}
initdataCopy2GPU(c2g_data, dst_roi_.height, dst_roi_.width);
}
if(flag == STITCH_SUCCESS)
{
LOGLN("\t~Prepare complete");
is_prepared_ = true;
}
return flag;
}
/*
* APAP算法的初始化
*/
int MyVideoStitcher::PrepareAPAP(vector<Mat> &src)
{
int num_images = static_cast<int>(src.size());
this->InitMembers(num_images);
// 计算一些放缩的尺度，在特征检测和计算接缝的时候，为了提高程序效率，可以对源图像进行一些放缩
work_megapix_ = -1; // 先不考虑放缩
seam_megapix_ = -1; // 先不考虑放缩
this->SetScales(src);
// 特征检测
vector<ImageFeatures> features(num_images);
this->FindFeatures(src, features);
// 特征匹配，并去掉噪声图片
vector<MatchesInfo> pairwise_matches;
this->MatchImages(features, pairwise_matches);
// APAP算法
APAPWarper apap_warper;
apap_warper.buildMaps(src, features, pairwise_matches, xmaps_, ymaps_, corners_);
for(int i = 0; i < num_images; i++)
sizes_[i] = xmaps_[i].size();
dst_roi_ = resultRoi(corners_, sizes_);
// 计算接缝
vector<Mat> seamed_masks(num_images);
vector<Mat> images_warped(num_images);
vector<Mat> init_masks(num_images);
for(int i = 0; i < num_images; i++)
{
init_masks[i].create(src[i].size(), CV_8U);
init_masks[i].setTo(Scalar::all(255));
remap(src[i], images_warped[i], xmaps_[i], ymaps_[i], INTER_LINEAR);
remap(init_masks[i], final_warped_masks_[i], xmaps_[i], ymaps_[i], INTER_NEAREST, BORDER_CONSTANT);
seamed_masks[i] = final_warped_masks_[i].clone();
}
this->FindSeam(images_warped, seamed_masks);
LOGLN("find seam");
// 曝光补偿
compensator_.createWeightMaps(corners_, images_warped, final_warped_masks_, ec_weight_maps_);
// 曝光补偿时，各像素权值也要resize一下
compensator_.gainMapResize(sizes_, ec_weight_maps_);
LOGLN("compensate");
images_warped.clear();
// 计算融合时，各像素的权值
Size dst_sz = dst_roi_.size();
//cout << "dst size: " << dst_sz << endl;
float blend_width = sqrt(static_cast<float>(dst_sz.area())) * blend_strength_ / 100.f;
blender_.setSharpness(1.f / blend_width);
for(int i = 0; i < num_images; i++)
final_blend_masks_[i] = final_warped_masks_[i] & seamed_masks[i];
blender_.createWeightMaps(dst_roi_, corners_, seamed_masks, blend_weight_maps_);