圖像拼接（image stitching）

時間 2019-11-06

標籤圖像拼接 image stitching 简体版

原文原文鏈接

# OpenCV中stitching的使用ios

OpenCV提供了高級別的函數封裝在Stitcher類中，使用很方便，不用考慮太多的細節。git

低級別函數封裝在detail命名空間中，展現了OpenCV算法實現的不少步驟和細節，使熟悉以下拼接流水線的用戶，方便本身定製。github

可見OpenCV圖像拼接模塊的實現是十分精密和複雜的，拼接的結果很完善，但同時也是費時的，徹底不可以實現實時應用。算法

官方提供的stitching和stitching_detailed使用示例，分別是高級別和低級別封裝這兩種方式正確地使用示例。兩種結果產生的拼接結果相同，後者卻能夠容許用戶，在參數變量初始化時，選擇各項算法。api

具體算法流程：app

命令行調用程序，輸入源圖像以及程序的參數
特徵點檢測，判斷是使用surf仍是orb，默認是surf。
對圖像的特徵點進行匹配，使用最近鄰和次近鄰方法，將兩個最優的匹配的置信度保存下來。
對圖像進行排序以及將置信度高的圖像保存到同一個集合中，刪除置信度比較低的圖像間的匹配，獲得能正確匹配的圖像序列。這樣將置信度高於門限的全部匹配合併到一個集合中。
對全部圖像進行相機參數粗略估計，而後求出旋轉矩陣
使用光束平均法進一步精準的估計出旋轉矩陣。
波形校訂，水平或者垂直
拼接
融合，多頻段融合，光照補償，

代碼：ide

#include "pch.h"
#include <iostream>
#include <fstream>
#include <string>
#include "opencv2/opencv_modules.hpp"
#include <opencv2/core/utility.hpp>
#include "opencv2/imgcodecs.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/stitching/detail/autocalib.hpp"
#include "opencv2/stitching/detail/blenders.hpp"
#include "opencv2/stitching/detail/timelapsers.hpp"
#include "opencv2/stitching/detail/camera.hpp"
#include "opencv2/stitching/detail/exposure_compensate.hpp"
#include "opencv2/stitching/detail/matchers.hpp"
#include "opencv2/stitching/detail/motion_estimators.hpp"
#include "opencv2/stitching/detail/seam_finders.hpp"
#include "opencv2/stitching/detail/warpers.hpp"
#include "opencv2/stitching/warpers.hpp"

#ifdef HAVE_OPENCV_XFEATURES2D
#include "opencv2/xfeatures2d/nonfree.hpp"
#endif

#define ENABLE_LOG 1
#define LOG(msg) std::cout << msg
#define LOGLN(msg) std::cout << msg << std::endl

using namespace std;
using namespace cv;
using namespace cv::detail;

static void printUsage()
{
	cout <<
		"Rotation model images stitcher.\n\n"
		"stitching_detailed img1 img2 [...imgN] [flags]\n\n"
		"Flags:\n"
		"  --preview\n"
		"      Run stitching in the preview mode. Works faster than usual mode,\n"
		"      but output image will have lower resolution.\n"
		"  --try_cuda (yes|no)\n"
		"      Try to use CUDA. The default value is 'no'. All default values\n"
		"      are for CPU mode.\n"
		"\nMotion Estimation Flags:\n"
		"  --work_megapix <float>\n"
		"      Resolution for image registration step. The default is 0.6 Mpx.\n"
		"  --features (surf|orb|sift|akaze)\n"
		"      Type of features used for images matching.\n"
		"      The default is surf if available, orb otherwise.\n"
		"  --matcher (homography|affine)\n"
		"      Matcher used for pairwise image matching.\n"
		"  --estimator (homography|affine)\n"
		"      Type of estimator used for transformation estimation.\n"
		"  --match_conf <float>\n"
		"      Confidence for feature matching step. The default is 0.65 for surf and 0.3 for orb.\n"
		"  --conf_thresh <float>\n"
		"      Threshold for two images are from the same panorama confidence.\n"
		"      The default is 1.0.\n"
		"  --ba (no|reproj|ray|affine)\n"
		"      Bundle adjustment cost function. The default is ray.\n"
		"  --ba_refine_mask (mask)\n"
		"      Set refinement mask for bundle adjustment. It looks like 'x_xxx',\n"
		"      where 'x' means refine respective parameter and '_' means don't\n"
		"      refine one, and has the following format:\n"
		"      <fx><skew><ppx><aspect><ppy>. The default mask is 'xxxxx'. If bundle\n"
		"      adjustment doesn't support estimation of selected parameter then\n"
		"      the respective flag is ignored.\n"
		"  --wave_correct (no|horiz|vert)\n"
		"      Perform wave effect correction. The default is 'horiz'.\n"
		"  --save_graph <file_name>\n"
		"      Save matches graph represented in DOT language to <file_name> file.\n"
		"      Labels description: Nm is number of matches, Ni is number of inliers,\n"
		"      C is confidence.\n"
		"\nCompositing Flags:\n"
		"  --warp (affine|plane|cylindrical|spherical|fisheye|stereographic|compressedPlaneA2B1|compressedPlaneA1.5B1|compressedPlanePortraitA2B1|compressedPlanePortraitA1.5B1|paniniA2B1|paniniA1.5B1|paniniPortraitA2B1|paniniPortraitA1.5B1|mercator|transverseMercator)\n"
		"      Warp surface type. The default is 'spherical'.\n"
		"  --seam_megapix <float>\n"
		"      Resolution for seam estimation step. The default is 0.1 Mpx.\n"
		"  --seam (no|voronoi|gc_color|gc_colorgrad)\n"
		"      Seam estimation method. The default is 'gc_color'.\n"
		"  --compose_megapix <float>\n"
		"      Resolution for compositing step. Use -1 for original resolution.\n"
		"      The default is -1.\n"
		"  --expos_comp (no|gain|gain_blocks|channels|channels_blocks)\n"
		"      Exposure compensation method. The default is 'gain_blocks'.\n"
		"  --expos_comp_nr_feeds <int>\n"
		"      Number of exposure compensation feed. The default is 1.\n"
		"  --expos_comp_nr_filtering <int>\n"
		"      Number of filtering iterations of the exposure compensation gains.\n"
		"      Only used when using a block exposure compensation method.\n"
		"      The default is 2.\n"
		"  --expos_comp_block_size <int>\n"
		"      BLock size in pixels used by the exposure compensator.\n"
		"      Only used when using a block exposure compensation method.\n"
		"      The default is 32.\n"
		"  --blend (no|feather|multiband)\n"
		"      Blending method. The default is 'multiband'.\n"
		"  --blend_strength <float>\n"
		"      Blending strength from [0,100] range. The default is 5.\n"
		"  --output <result_img>\n"
		"      The default is 'result.jpg'.\n"
		"  --timelapse (as_is|crop) \n"
		"      Output warped images separately as frames of a time lapse movie, with 'fixed_' prepended to input file names.\n"
		"  --rangewidth <int>\n"
		"      uses range_width to limit number of images to match with.\n";
}


// Default command line args
vector<String> img_names;
bool preview = false;
bool try_cuda = false;
double work_megapix = 0.6;
double seam_megapix = 0.1;
double compose_megapix = -1;
float conf_thresh = 1.f;
#ifdef HAVE_OPENCV_XFEATURES2D
string features_type = "surf";
#else
string features_type = "orb";
#endif
string matcher_type = "homography";
string estimator_type = "homography";
string ba_cost_func = "ray";
string ba_refine_mask = "xxxxx";
bool do_wave_correct = true;
WaveCorrectKind wave_correct = detail::WAVE_CORRECT_HORIZ;
bool save_graph = false;
std::string save_graph_to;
string warp_type = "spherical";
int expos_comp_type = ExposureCompensator::GAIN_BLOCKS;
int expos_comp_nr_feeds = 1;
int expos_comp_nr_filtering = 2;
int expos_comp_block_size = 32;
float match_conf = 0.3f;
string seam_find_type = "gc_color";
int blend_type = Blender::MULTI_BAND;
int timelapse_type = Timelapser::AS_IS;
float blend_strength = 5;
string result_name = "F:/opencv/build/bin/sample-data/stitching/result.jpg";
bool timelapse = false;
int range_width = -1;


int main(int argc, char* argv[])
{
#if ENABLE_LOG
	int64 app_start_time = getTickCount();
#endif

#if 0
	cv::setBreakOnError(true);
#endif

	img_names.push_back("F:/opencv/build/bin/sample-data/stitching/st1.jpg");
	img_names.push_back("F:/opencv/build/bin/sample-data/stitching/st2.jpg");
	img_names.push_back("F:/opencv/build/bin/sample-data/stitching/st3.jpg");
	img_names.push_back("F:/opencv/build/bin/sample-data/stitching/st4.jpg");

	// Check if have enough images
	int num_images = static_cast<int>(img_names.size());
	if (num_images < 2)
	{
		LOGLN("Need more images");
		return -1;
	}

	double work_scale = 1, seam_scale = 1, compose_scale = 1;
	bool is_work_scale_set = false, is_seam_scale_set = false, is_compose_scale_set = false;

	LOGLN("Finding features...");
#if ENABLE_LOG
	int64 t = getTickCount();
#endif

	Ptr<Feature2D> finder;
	if (features_type == "orb")
	{
		finder = ORB::create();
	}
	else if (features_type == "akaze")
	{
		finder = AKAZE::create();
	}
#ifdef HAVE_OPENCV_XFEATURES2D
	else if (features_type == "surf")
	{
		finder = xfeatures2d::SURF::create();
	}
	else if (features_type == "sift") {
		finder = xfeatures2d::SIFT::create();
	}
#endif
	else
	{
		cout << "Unknown 2D features type: '" << features_type << "'.\n";
		return -1;
	}

	cout << "Current 2D features type: '" << features_type << "'.\n";

	Mat full_img, img;
	vector<ImageFeatures> features(num_images);
	vector<Mat> images(num_images);
	vector<Size> full_img_sizes(num_images);
	double seam_work_aspect = 1;

	for (int i = 0; i < num_images; ++i)
	{
		full_img = imread(samples::findFile(img_names[i]));
		full_img_sizes[i] = full_img.size();

		if (full_img.empty())
		{
			LOGLN("Can't open image " << img_names[i]);
			return -1;
		}
		if (work_megapix < 0)
		{
			img = full_img;
			work_scale = 1;
			is_work_scale_set = true;
		}
		else
		{
			if (!is_work_scale_set)
			{
				work_scale = min(1.0, sqrt(work_megapix * 1e6 / full_img.size().area()));
				is_work_scale_set = true;
			}
			resize(full_img, img, Size(), work_scale, work_scale, INTER_LINEAR_EXACT);
		}
		if (!is_seam_scale_set)
		{
			seam_scale = min(1.0, sqrt(seam_megapix * 1e6 / full_img.size().area()));
			seam_work_aspect = seam_scale / work_scale;
			is_seam_scale_set = true;
		}

		computeImageFeatures(finder, img, features[i]);
		features[i].img_idx = i;
		LOGLN("Features in image #" << i + 1 << ": " << features[i].keypoints.size());

		resize(full_img, img, Size(), seam_scale, seam_scale, INTER_LINEAR_EXACT);
		images[i] = img.clone();
	}

	full_img.release();
	img.release();

	LOGLN("Finding features, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");

	LOG("Pairwise matching");
#if ENABLE_LOG
	t = getTickCount();
#endif
	vector<MatchesInfo> pairwise_matches;
	Ptr<FeaturesMatcher> matcher;
	if (matcher_type == "affine")
		matcher = makePtr<AffineBestOf2NearestMatcher>(false, try_cuda, match_conf);
	else if (range_width == -1)
		matcher = makePtr<BestOf2NearestMatcher>(try_cuda, match_conf);
	else
		matcher = makePtr<BestOf2NearestRangeMatcher>(range_width, try_cuda, match_conf);

	(*matcher)(features, pairwise_matches);
	matcher->collectGarbage();

	LOGLN("Pairwise matching, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");

	// Check if we should save matches graph
	if (save_graph)
	{
		LOGLN("Saving matches graph...");
		ofstream f(save_graph_to.c_str());
		f << matchesGraphAsString(img_names, pairwise_matches, conf_thresh);
	}

	// Leave only images we are sure are from the same panorama
	vector<int> indices = leaveBiggestComponent(features, pairwise_matches, conf_thresh);
	vector<Mat> img_subset;
	vector<String> img_names_subset;
	vector<Size> full_img_sizes_subset;
	for (size_t i = 0; i < indices.size(); ++i)
	{
		img_names_subset.push_back(img_names[indices[i]]);
		img_subset.push_back(images[indices[i]]);
		full_img_sizes_subset.push_back(full_img_sizes[indices[i]]);
	}

	images = img_subset;
	img_names = img_names_subset;
	full_img_sizes = full_img_sizes_subset;

	// Check if we still have enough images
	num_images = static_cast<int>(img_names.size());
	if (num_images < 2)
	{
		LOGLN("Need more images from the same panorama");
		return -1;
	}

	Ptr<Estimator> estimator;
	if (estimator_type == "affine")
		estimator = makePtr<AffineBasedEstimator>();
	else
		estimator = makePtr<HomographyBasedEstimator>();

	vector<CameraParams> cameras;
	if (!(*estimator)(features, pairwise_matches, cameras))
	{
		cout << "Homography estimation failed.\n";
		return -1;
	}

	for (size_t i = 0; i < cameras.size(); ++i)
	{
		Mat R;
		cameras[i].R.convertTo(R, CV_32F);
		cameras[i].R = R;
		LOGLN("Initial camera intrinsics #" << indices[i] + 1 << ":\nK:\n" << cameras[i].K() << "\nR:\n" << cameras[i].R);
	}

	Ptr<detail::BundleAdjusterBase> adjuster;
	if (ba_cost_func == "reproj") adjuster = makePtr<detail::BundleAdjusterReproj>();
	else if (ba_cost_func == "ray") adjuster = makePtr<detail::BundleAdjusterRay>();
	else if (ba_cost_func == "affine") adjuster = makePtr<detail::BundleAdjusterAffinePartial>();
	else if (ba_cost_func == "no") adjuster = makePtr<NoBundleAdjuster>();
	else
	{
		cout << "Unknown bundle adjustment cost function: '" << ba_cost_func << "'.\n";
		return -1;
	}
	adjuster->setConfThresh(conf_thresh);
	Mat_<uchar> 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);
	if (!(*adjuster)(features, pairwise_matches, cameras))
	{
		cout << "Camera parameters adjusting failed.\n";
		return -1;
	}

	// Find median focal length

	vector<double> focals;
	for (size_t i = 0; i < cameras.size(); ++i)
	{
		LOGLN("Camera #" << indices[i] + 1 << ":\nK:\n" << cameras[i].K() << "\nR:\n" << cameras[i].R);
		focals.push_back(cameras[i].focal);
	}

	sort(focals.begin(), focals.end());
	float warped_image_scale;
	if (focals.size() % 2 == 1)
		warped_image_scale = static_cast<float>(focals[focals.size() / 2]);
	else
		warped_image_scale = static_cast<float>(focals[focals.size() / 2 - 1] + focals[focals.size() / 2]) * 0.5f;

	if (do_wave_correct)
	{
		vector<Mat> rmats;
		for (size_t i = 0; i < cameras.size(); ++i)
			rmats.push_back(cameras[i].R.clone());
		waveCorrect(rmats, wave_correct);
		for (size_t i = 0; i < cameras.size(); ++i)
			cameras[i].R = rmats[i];
	}

	LOGLN("Warping images (auxiliary)... ");
#if ENABLE_LOG
	t = getTickCount();
#endif

	vector<Point> corners(num_images);
	vector<UMat> masks_warped(num_images);
	vector<UMat> images_warped(num_images);
	vector<Size> sizes(num_images);
	vector<UMat> masks(num_images);

	// Preapre images masks
	for (int i = 0; i < num_images; ++i)
	{
		masks[i].create(images[i].size(), CV_8U);
		masks[i].setTo(Scalar::all(255));
	}

	// Warp images and their masks

	Ptr<WarperCreator> warper_creator;
#ifdef HAVE_OPENCV_CUDAWARPING
	if (try_cuda && cuda::getCudaEnabledDeviceCount() > 0)
	{
		if (warp_type == "plane")
			warper_creator = makePtr<cv::PlaneWarperGpu>();
		else if (warp_type == "cylindrical")
			warper_creator = makePtr<cv::CylindricalWarperGpu>();
		else if (warp_type == "spherical")
			warper_creator = makePtr<cv::SphericalWarperGpu>();
	}
	else
#endif
	{
		if (warp_type == "plane")
			warper_creator = makePtr<cv::PlaneWarper>();
		else if (warp_type == "affine")
			warper_creator = makePtr<cv::AffineWarper>();
		else if (warp_type == "cylindrical")
			warper_creator = makePtr<cv::CylindricalWarper>();
		else if (warp_type == "spherical")
			warper_creator = makePtr<cv::SphericalWarper>();
		else if (warp_type == "fisheye")
			warper_creator = makePtr<cv::FisheyeWarper>();
		else if (warp_type == "stereographic")
			warper_creator = makePtr<cv::StereographicWarper>();
		else if (warp_type == "compressedPlaneA2B1")
			warper_creator = makePtr<cv::CompressedRectilinearWarper>(2.0f, 1.0f);
		else if (warp_type == "compressedPlaneA1.5B1")
			warper_creator = makePtr<cv::CompressedRectilinearWarper>(1.5f, 1.0f);
		else if (warp_type == "compressedPlanePortraitA2B1")
			warper_creator = makePtr<cv::CompressedRectilinearPortraitWarper>(2.0f, 1.0f);
		else if (warp_type == "compressedPlanePortraitA1.5B1")
			warper_creator = makePtr<cv::CompressedRectilinearPortraitWarper>(1.5f, 1.0f);
		else if (warp_type == "paniniA2B1")
			warper_creator = makePtr<cv::PaniniWarper>(2.0f, 1.0f);
		else if (warp_type == "paniniA1.5B1")
			warper_creator = makePtr<cv::PaniniWarper>(1.5f, 1.0f);
		else if (warp_type == "paniniPortraitA2B1")
			warper_creator = makePtr<cv::PaniniPortraitWarper>(2.0f, 1.0f);
		else if (warp_type == "paniniPortraitA1.5B1")
			warper_creator = makePtr<cv::PaniniPortraitWarper>(1.5f, 1.0f);
		else if (warp_type == "mercator")
			warper_creator = makePtr<cv::MercatorWarper>();
		else if (warp_type == "transverseMercator")
			warper_creator = makePtr<cv::TransverseMercatorWarper>();
	}

	if (!warper_creator)
	{
		cout << "Can't create the following warper '" << warp_type << "'\n";
		return 1;
	}

	Ptr<RotationWarper> warper = warper_creator->create(static_cast<float>(warped_image_scale * seam_work_aspect));

	for (int i = 0; i < num_images; ++i)
	{
		Mat_<float> K;
		cameras[i].K().convertTo(K, CV_32F);
		float swa = (float)seam_work_aspect;
		K(0, 0) *= swa; K(0, 2) *= swa;
		K(1, 1) *= swa; K(1, 2) *= swa;

		corners[i] = warper->warp(images[i], K, cameras[i].R, INTER_LINEAR, BORDER_REFLECT, images_warped[i]);
		sizes[i] = images_warped[i].size();

		warper->warp(masks[i], K, cameras[i].R, INTER_NEAREST, BORDER_CONSTANT, masks_warped[i]);
	}

	vector<UMat> images_warped_f(num_images);
	for (int i = 0; i < num_images; ++i)
		images_warped[i].convertTo(images_warped_f[i], CV_32F);

	LOGLN("Warping images, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");

	LOGLN("Compensating exposure...");
#if ENABLE_LOG
	t = getTickCount();
#endif

	Ptr<ExposureCompensator> compensator = ExposureCompensator::createDefault(expos_comp_type);
	if (dynamic_cast<GainCompensator*>(compensator.get()))
	{
		GainCompensator* gcompensator = dynamic_cast<GainCompensator*>(compensator.get());
		gcompensator->setNrFeeds(expos_comp_nr_feeds);
	}

	if (dynamic_cast<ChannelsCompensator*>(compensator.get()))
	{
		ChannelsCompensator* ccompensator = dynamic_cast<ChannelsCompensator*>(compensator.get());
		ccompensator->setNrFeeds(expos_comp_nr_feeds);
	}

	if (dynamic_cast<BlocksCompensator*>(compensator.get()))
	{
		BlocksCompensator* bcompensator = dynamic_cast<BlocksCompensator*>(compensator.get());
		bcompensator->setNrFeeds(expos_comp_nr_feeds);
		bcompensator->setNrGainsFilteringIterations(expos_comp_nr_filtering);
		bcompensator->setBlockSize(expos_comp_block_size, expos_comp_block_size);
	}

	compensator->feed(corners, images_warped, masks_warped);

	LOGLN("Compensating exposure, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");

	LOGLN("Finding seams...");
#if ENABLE_LOG
	t = getTickCount();
#endif

	Ptr<SeamFinder> seam_finder;
	if (seam_find_type == "no")
		seam_finder = makePtr<detail::NoSeamFinder>();
	else if (seam_find_type == "voronoi")
		seam_finder = makePtr<detail::VoronoiSeamFinder>();
	else if (seam_find_type == "gc_color")
	{
#ifdef HAVE_OPENCV_CUDALEGACY
		if (try_cuda && cuda::getCudaEnabledDeviceCount() > 0)
			seam_finder = makePtr<detail::GraphCutSeamFinderGpu>(GraphCutSeamFinderBase::COST_COLOR);
		else
#endif
			seam_finder = makePtr<detail::GraphCutSeamFinder>(GraphCutSeamFinderBase::COST_COLOR);
	}
	else if (seam_find_type == "gc_colorgrad")
	{
#ifdef HAVE_OPENCV_CUDALEGACY
		if (try_cuda && cuda::getCudaEnabledDeviceCount() > 0)
			seam_finder = makePtr<detail::GraphCutSeamFinderGpu>(GraphCutSeamFinderBase::COST_COLOR_GRAD);
		else
#endif
			seam_finder = makePtr<detail::GraphCutSeamFinder>(GraphCutSeamFinderBase::COST_COLOR_GRAD);
	}
	else if (seam_find_type == "dp_color")
		seam_finder = makePtr<detail::DpSeamFinder>(DpSeamFinder::COLOR);
	else if (seam_find_type == "dp_colorgrad")
		seam_finder = makePtr<detail::DpSeamFinder>(DpSeamFinder::COLOR_GRAD);
	if (!seam_finder)
	{
		cout << "Can't create the following seam finder '" << seam_find_type << "'\n";
		return 1;
	}

	seam_finder->find(images_warped_f, corners, masks_warped);

	LOGLN("Finding seams, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");

	// Release unused memory
	images.clear();
	images_warped.clear();
	images_warped_f.clear();
	masks.clear();

	LOGLN("Compositing...");
#if ENABLE_LOG
	t = getTickCount();
#endif

	Mat img_warped, img_warped_s;
	Mat dilated_mask, seam_mask, mask, mask_warped;
	Ptr<Blender> blender;
	Ptr<Timelapser> timelapser;
	//double compose_seam_aspect = 1;
	double compose_work_aspect = 1;

	for (int img_idx = 0; img_idx < num_images; ++img_idx)
	{
		LOGLN("Compositing image #" << indices[img_idx] + 1);

		// Read image and resize it if necessary
		full_img = imread(samples::findFile(img_names[img_idx]));
		if (!is_compose_scale_set)
		{
			if (compose_megapix > 0)
				compose_scale = min(1.0, sqrt(compose_megapix * 1e6 / full_img.size().area()));
			is_compose_scale_set = true;

			// Compute relative scales
			//compose_seam_aspect = compose_scale / seam_scale;
			compose_work_aspect = compose_scale / work_scale;

			// Update warped image scale
			warped_image_scale *= static_cast<float>(compose_work_aspect);
			warper = warper_creator->create(warped_image_scale);

			// Update corners and sizes
			for (int i = 0; i < num_images; ++i)
			{
				// Update intrinsics
				cameras[i].focal *= compose_work_aspect;
				cameras[i].ppx *= compose_work_aspect;
				cameras[i].ppy *= compose_work_aspect;

				// Update corner and size
				Size sz = full_img_sizes[i];
				if (std::abs(compose_scale - 1) > 1e-1)
				{
					sz.width = cvRound(full_img_sizes[i].width * compose_scale);
					sz.height = cvRound(full_img_sizes[i].height * compose_scale);
				}

				Mat K;
				cameras[i].K().convertTo(K, CV_32F);
				Rect roi = warper->warpRoi(sz, K, cameras[i].R);
				corners[i] = roi.tl();
				sizes[i] = roi.size();
			}
		}
		if (abs(compose_scale - 1) > 1e-1)
			resize(full_img, img, Size(), compose_scale, compose_scale, INTER_LINEAR_EXACT);
		else
			img = full_img;
		full_img.release();
		Size img_size = img.size();

		Mat K;
		cameras[img_idx].K().convertTo(K, CV_32F);

		// Warp the current image
		warper->warp(img, K, cameras[img_idx].R, INTER_LINEAR, BORDER_REFLECT, img_warped);

		// Warp the current image mask
		mask.create(img_size, CV_8U);
		mask.setTo(Scalar::all(255));
		warper->warp(mask, K, cameras[img_idx].R, INTER_NEAREST, BORDER_CONSTANT, mask_warped);

		// Compensate exposure
		compensator->apply(img_idx, corners[img_idx], img_warped, mask_warped);

		img_warped.convertTo(img_warped_s, CV_16S);
		img_warped.release();
		img.release();
		mask.release();

		dilate(masks_warped[img_idx], dilated_mask, Mat());
		resize(dilated_mask, seam_mask, mask_warped.size(), 0, 0, INTER_LINEAR_EXACT);
		mask_warped = seam_mask & mask_warped;

		if (!blender && !timelapse)
		{
			blender = Blender::createDefault(blend_type, try_cuda);
			Size dst_sz = resultRoi(corners, sizes).size();
			float blend_width = sqrt(static_cast<float>(dst_sz.area())) * blend_strength / 100.f;
			if (blend_width < 1.f)
				blender = Blender::createDefault(Blender::NO, try_cuda);
			else if (blend_type == Blender::MULTI_BAND)
			{
				MultiBandBlender* mb = dynamic_cast<MultiBandBlender*>(blender.get());
				mb->setNumBands(static_cast<int>(ceil(log(blend_width) / log(2.)) - 1.));
				LOGLN("Multi-band blender, number of bands: " << mb->numBands());
			}
			else if (blend_type == Blender::FEATHER)
			{
				FeatherBlender* fb = dynamic_cast<FeatherBlender*>(blender.get());
				fb->setSharpness(1.f / blend_width);
				LOGLN("Feather blender, sharpness: " << fb->sharpness());
			}
			blender->prepare(corners, sizes);
		}
		else if (!timelapser && timelapse)
		{
			timelapser = Timelapser::createDefault(timelapse_type);
			timelapser->initialize(corners, sizes);
		}

		// Blend the current image
		if (timelapse)
		{
			timelapser->process(img_warped_s, Mat::ones(img_warped_s.size(), CV_8UC1), corners[img_idx]);
			String fixedFileName;
			size_t pos_s = String(img_names[img_idx]).find_last_of("/\\");
			if (pos_s == String::npos)
			{
				fixedFileName = "fixed_" + img_names[img_idx];
			}
			else
			{
				fixedFileName = "fixed_" + String(img_names[img_idx]).substr(pos_s + 1, String(img_names[img_idx]).length() - pos_s);
			}
			imwrite(fixedFileName, timelapser->getDst());
		}
		else
		{
			blender->feed(img_warped_s, mask_warped, corners[img_idx]);
		}
	}

	if (!timelapse)
	{
		Mat result, result_mask;
		blender->blend(result, result_mask);

		LOGLN("Compositing, time: " << ((getTickCount() - t) / getTickFrequency()) << " sec");

		imwrite(result_name, result);
	}

	LOGLN("Finished, total time: " << ((getTickCount() - app_start_time) / getTickFrequency()) << " sec");
	return 0;
}

結果：函數

Finding features...
Current 2D features type: 'surf'.
[ INFO:0] Initialize OpenCL runtime...
Features in image #1: 911
Features in image #2: 1085
Features in image #3: 1766
Features in image #4: 2001
Finding features, time: 3.33727 sec
Pairwise matchingPairwise matching, time: 3.2849 sec
Initial camera intrinsics #1:
K:
[4503.939581818162, 0, 285;
 0, 4503.939581818162, 210;
 0, 0, 1]
R:
[1.0011346, 0.0019526235, -0.0037489906;
 0.00011878588, 1.0000151, -0.052518897;
 -0.0011389133, 0.021224562, 1]
Initial camera intrinsics #2:
K:
[4503.939581818162, 0, 249;
 0, 4503.939581818162, 222;
 0, 0, 1]
R:
[1.0023992, 0.0045258515, 0.083801955;
 -9.7107059e-06, 1.0006112, -0.049870808;
 0.015923418, 0.048128795, 1.0000379]
Initial camera intrinsics #3:
K:
[4503.939581818162, 0, 302.5;
 0, 4503.939581818162, 173.5;
 0, 0, 1]
R:
[1, 0, 0;
 0, 1, 0;
 0, 0, 1]
Initial camera intrinsics #4:
K:
[4503.939581818162, 0, 274.5;
 0, 4503.939581818162, 194.5;
 0, 0, 1]
R:
[1.0004042, 0.00080040237, 0.078620218;
 0.00026136645, 1.0005095, -0.0048735617;
 0.0061902963, 0.0096427174, 1.0004393]
Camera #1:
K:
[6569.821976030652, 0, 285;
 0, 6569.821976030652, 210;
 0, 0, 1]
R:
[0.99999672, 0.00038595949, -0.0025201384;
 -0.00047636221, 0.99935275, -0.035969362;
 0.0025046244, 0.035970442, 0.99934971]
Camera #2:
K:
[6571.327169846625, 0, 249;
 0, 6571.327169846625, 222;
 0, 0, 1]
R:
[0.99835128, 0.0012797765, 0.057385404;
 0.00068109832, 0.99941689, -0.03413773;
 -0.05739563, 0.034120534, 0.99776828]
Camera #3:
K:
[6570.486320822205, 0, 302.5;
 0, 6570.486320822205, 173.5;
 0, 0, 1]
R:
[1, -1.2951205e-09, 0;
 -1.2914825e-09, 1, 0;
 0, -4.6566129e-10, 1]
Camera #4:
K:
[6571.394840241929, 0, 274.5;
 0, 6571.394840241929, 194.5;
 0, 0, 1]
R:
[0.99855018, -0.00018820527, 0.053829439;
 0.0003683792, 0.99999434, -0.0033372282;
 -0.053828511, 0.0033522192, 0.99854457]
Warping images (auxiliary)...
[ INFO:0] Successfully initialized OpenCL cache directory: C:\Users\mzhu\AppData\Local\Temp\opencv\4.1\opencl_cache\
[ INFO:0] Preparing OpenCL cache configuration for context: Intel_R__Corporation--Intel_R__HD_Graphics_620--21_20_16_4574
Warping images, time: 0.0817463 sec
Compensating exposure...
Compensating exposure, time: 0.22982 sec
Finding seams...
Finding seams, time: 1.49795 sec
Compositing...
Compositing image #1
Multi-band blender, number of bands: 5
Compositing image #2
Compositing image #3
Compositing image #4
Compositing, time: 0.705931 sec
Finished, total time: 116.51 sec

# OpenPano：如何編寫一個全景拼接器post

OpenPano: Automatic Panorama Stitching From Scratch （https://github.com/ppwwyyxx/OpenPano）ui

StitchIt: Optimization and Parallelization of Image Stitching （https://github.com/stitchit/StitchIt）

ParaPano: Parallel image stitching using CUDA （https://github.com/zq-chen/ParaPano）

NISwGSP : Natural Image Stitching with the Global Similarity Prior （For Windows:https://github.com/firdauslubis88/NISwGSP，論文閱讀筆記：https://zhuanlan.zhihu.com/p/57543736）