opencv----计算图像旋转参数

最新推荐文章于 2024-08-14 20:29:59 发布
原创最新推荐文章于 2024-08-14 20:29:59 发布 · 1.1k 阅读
1 ·
CC 4.0 BY-SA版权
文章标签：
#fourier #opencv #图像处理
opencv 专栏收录该内容
23 篇文章
订阅专栏
本文介绍了一种基于图像傅里叶变换的方法来识别两张图像之间的平移和旋转差异。通过计算图像的傅里叶变换、功率谱及对数极坐标变换等步骤，实现了图像匹配中的平移与旋转识别。
摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >
弄这个弄了好几天，一直想的复杂了
其实很简单，其实并不难
先上代码吧，不过里面有很多多余的函数，不过之后计算平移+旋转的时候会用到
// xuanzhuan.cpp : 定义控制台应用程序的入口点。
//

#include "StdAfx.h"



#include <opencv2/opencv.hpp>
#include <opencv2/core/core.hpp>
#include <opencv2/features2d/features2d.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/legacy/legacy.hpp>
#include <iostream>
#include <vector>

using namespace cv;
using namespace std;

void fourier(cv::Mat in,cv::Mat &out_re,cv::Mat &out_im);//计算傅里叶变换，分别输出变换后结果的实部和虚部
void gonglvpu(cv::Mat in1,cv::Mat in2,cv::Mat out);//计算傅里叶变换后的功率谱
void pingyi(cv::Mat in_1,cv::Mat in_2,int &x,int &y);
void cart2polar(cv::Mat in,cv::Mat &out);
void polar2cart(cv::Mat in,cv::Mat out);
void fft2(IplImage *src, IplImage *dst);
void fft2shift(IplImage *src, IplImage *dst);
void test(cv::Mat in,cv::Mat out);//本函数可以实现计算图像傅里叶变换后功率谱的功能
int _tmain(int argc, _TCHAR* argv[])
{

	cv::Mat image1=cv::imread("lena.bmp");
	cv::Mat image2=cv::imread("5du.bmp");
	cv::cvtColor(image1,image1,CV_RGB2GRAY);
	cv::cvtColor(image2,image2,CV_RGB2GRAY);
	cart2polar(image1,image1);
	//image1=image1(cv::Rect(0,0,200,200));
	cart2polar(image2,image2);
	//image2=image2(cv::Rect(0,0,200,200));
	cv::namedWindow("www");
	cv::imshow("www",image2);
	int xx=0,yy=0;
	pingyi(image1,image2,xx,yy);
	cout<<xx<<"     "<<yy<<endl;
	int theta=((double)(yy+1)/image1.rows)*360;
	cout<<"theta  "<<theta<<endl;
	cv::waitKey();
	/*
	cv::Mat src1s[]={Mat_<float>(image1),cv::Mat::zeros(image1.size(),CV_32FC1)};
	cv::Mat FA1;
	cv::merge(src1s,2,FA1);
	//cv::dft(FA1,FA1);
	cv::split(FA1,src1s);
	cv::magnitude(src1s[0],src1s[1],src1s[0]);
	src1s[0]=src1s[0](cv::Rect(100,100,400,400));
	cart2polar(src1s[0],src1s[0]);
	//polar2cart(src1s[0],src1s[0]);
	src1s[0]=src1s[0](cv::Rect(0,0,200,200));
	cv::normalize(src1s[0],src1s[0],0,1,CV_MINMAX);

	cv::Mat src2s[]={Mat_<float>(image2),cv::Mat::zeros(image2.size(),CV_32FC1)};
	cv::Mat FA2;
	cv::merge(src2s,2,FA2);
	//cv::dft(FA2,FA2);
	cv::split(FA2,src2s);
	cv::magnitude(src2s[0],src2s[1],src2s[0]);
	src2s[0]=src2s[0](cv::Rect(100,100,400,400));
	cart2polar(src2s[0],src2s[0]);
	src2s[0]=src2s[0](cv::Rect(0,0,200,200));
	cv::normalize(src2s[0],src2s[0],0,1,CV_MINMAX);
	
	//cv::normalize(src2s[0],src2s[0],0,1,CV_MINMAX);
	cv::namedWindow("test");
	cv::imshow("test",src2s[0]);
	cv::waitKey();

	int xx=0,yy=0;
	pingyi(src1s[0],src2s[0],xx,yy);
	cout<<xx<<"     "<<yy<<endl;

		
	*/
	return 0;
}

void test(cv::Mat in,cv::Mat out)
{
	IplImage ss=in;
	IplImage tt=in;
	IplImage *src=&in.operator IplImage();          //源图像
	IplImage *Image=&out.operator IplImage();
	IplImage *Fourier;   //傅里叶系数
	IplImage *dst ;


	IplImage *ImageRe;
	IplImage *ImageIm;
	IplImage *ImageIm1;

	
	IplImage *ImageDst;


	double m,M;
	double scale;
	double shift;
	//src = cvLoadImage("lena.bmp",0);   //加载源图像，第二个参数表示将输入的图片转为单信道 
	Fourier = cvCreateImage(cvGetSize(src),IPL_DEPTH_64F,2);
	dst = cvCreateImage(cvGetSize(src),IPL_DEPTH_64F,2);
	ImageRe = cvCreateImage(cvGetSize(src),IPL_DEPTH_64F,1);
	ImageIm = cvCreateImage(cvGetSize(src),IPL_DEPTH_64F,1);
	ImageIm1 = cvCreateImage(cvGetSize(src),IPL_DEPTH_64F,1);
	//Image = cvCreateImage(cvGetSize(src),src->depth,src->nChannels);
	ImageDst = cvCreateImage(cvGetSize(src),src->depth,src->nChannels);
	fft2(src,Fourier);                  //傅里叶变换
	fft2shift(Fourier, Image);          //中心化
	//cvDFT(Fourier,dst,CV_DXT_INV_SCALE);//实现傅里叶逆变换，并对结果进行缩放
	
	cvSplit(dst,ImageRe,ImageIm,0,0);
	//cvNamedWindow("源图像",0);
	//cvShowImage("源图像",src);             
	//对数组每个元素平方并存储在第二个参数中
	cvPow(ImageRe,ImageRe,2);               
	cvPow(ImageIm,ImageIm1,2);
	cvAdd(ImageRe,ImageIm1,ImageRe,NULL);
	cvPow(ImageRe,ImageRe,0.5);
	cvMinMaxLoc(ImageRe,&m,&M,NULL,NULL);
	scale = 255/(M - m);
	shift = -m * scale;
	//将shift加在ImageRe各元素按比例缩放的结果上，存储为ImageDst
	cvConvertScale(ImageRe,ImageDst,scale,shift);
	cvNormalize(ImageIm,ImageIm, 0, 1, CV_MINMAX);
	//cvNamedWindow("傅里叶谱",0);
	//cvShowImage("傅里叶谱",Image);
	//cvNamedWindow("傅里叶逆变换",0);
	//cvShowImage("傅里叶逆变换",ImageDst);
	//cvNamedWindow("test",0);
	//cvShowImage("test",ImageIm);
	cvWaitKey(0);
	//cvReleaseImage(&src);
	//cvReleaseImage(&Image);
	cvReleaseImage(&ImageIm);
	cvReleaseImage(&ImageRe);
	cvReleaseImage(&Fourier);
	cvReleaseImage(&dst);
	cvReleaseImage(&ImageDst);
    //    cvDestroyAllWindows();
}
void fourier(cv::Mat in,cv::Mat &out_re,cv::Mat &out_im)
{
	Mat planes[] = {Mat_<float>(in), Mat::zeros(in.size(), CV_32FC1)};
	Mat complexI;
	merge(planes, 2, complexI); 
	cv::dft(complexI, complexI);
	cv::split(complexI, planes);
	out_re=planes[0];
	out_im=planes[1];

}
void gonglvpu(cv::Mat in1,cv::Mat in2,cv::Mat out)
{
	cv::magnitude(in1,in2,out);
	/*
	IplImage data_1=in1;
	IplImage data_2=in2;
	IplImage data_3=out;
	IplImage* pI1  = &in1.operator IplImage();
	IplImage* pI2  = &in2.operator IplImage();
	IplImage* pI3  = &out.operator IplImage();
	cvPow(pI1,pI1,2);
	cvPow(pI2,pI2,2);
	cvAdd(pI1,pI2,pI3);
	cvPow(pI3,pI3,0.5);
	*/
}
void pingyi(cv::Mat in_1,cv::Mat in_2,int &x,int &y)
{
	
	cv::Mat padded=in_1;
	Mat planes[] = {Mat_<float>(padded), Mat::zeros(padded.size(), CV_32F)};
	Mat complexI;
	merge(planes, 2, complexI); 
	cv::dft(complexI, complexI);
	cv::split(complexI, planes);

	//////////////////////////////////
	
	cv::Mat padded2=in_2;
	Mat planes2[] = {Mat_<float>(padded2), Mat::zeros(padded2.size(), CV_32F)};
	Mat complexI2;
	merge(planes2, 2, complexI2); 
	cv::dft(complexI2, complexI2);
	cv::split(complexI2, planes2);
	//cv::normalize(planes[1],planes[1], 0, 1, CV_MINMAX);
	////////////////////////////////////////////////////////
	//src1变换后实部 planes[0]   虚部planes[1]
	//src2变换后实部 planes2[0]  虚部planes2[1]


	//互功率谱存放矩阵 实部  planes3[0]   虚部planes3[1]
	//////////////////////////////////////////////////////////
	
	Mat planes3[] = {Mat::zeros(in_1.size(), CV_32F), Mat::zeros(in_1.size(), CV_32F)};

	for(int i=0;i<in_1.rows-1;i++)
	{
		for(int j=0;j<in_1.cols-1;j++)
		{
			double r1=planes[0].at<float>(j,i);//获取像素值
			double i1=planes[1].at<float>(j,i);
			double r2=planes2[0].at<float>(j,i);
			double i2=planes2[1].at<float>(j,i);

			double r3    =r1*r2+i1*i2;//计算互功率谱的值 
			double i3    =r1*i2-r2*i1;
			double abs   =sqrt((r3*r3)+(i3*i3));
			double r_exp =r3/abs;
			double i_exp =i3/abs;

			
			planes3[0].at<float>(j,i)=r_exp;
			planes3[1].at<float>(j,i)=i_exp;
		}
	}

	Mat complexI3;
	merge(planes3, 2, complexI3);
	cv::dft(complexI3,complexI3,CV_DXT_INV_SCALE);
	cv::split(complexI3,planes3);
	magnitude(planes3[0], planes3[1], planes3[0]);
	complexI3=planes3[0];
	complexI3 += Scalar::all(1); 
	log(complexI3, complexI3);
	cv::normalize(complexI3,complexI3, 0, 1, CV_MINMAX);

	double max=0,min=0;
	cv::Point minLoc,maxLoc;
	cv::minMaxLoc(complexI3,&min,&max,&minLoc,&maxLoc);
	//cv::imwrite("complexI3.bmp",complexI3);


	x=maxLoc.x;
	y=maxLoc.y;
	cout<<min<<"   "<<max<<endl;
	cout<<complexI3.at<float>(maxLoc)<<endl;
}
void cart2polar(cv::Mat in,cv::Mat &out)
{
	IplImage tmp1=in;
	IplImage* pI1  = &in.operator IplImage();
	IplImage tmp2=out;
	IplImage* pI2  = &out.operator IplImage();
	cvLogPolar( pI1, pI2, cvPoint2D32f(pI1->width/2,pI1->height/2), 40, CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS );
	//cvReleaseImage(&pI1);
	//cvReleaseImage(&pI2);
}
void polar2cart(cv::Mat in,cv::Mat out)
{
	IplImage tmp1=in;
	IplImage* pI1  = &in.operator IplImage();
	IplImage tmp2=out;
	IplImage* pI2  = &out.operator IplImage();
	cvLogPolar( pI1, pI2, cvPoint2D32f(pI1->width/2,pI1->height/2), 40, CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS+CV_WARP_INVERSE_MAP );
}
void fft2(IplImage *src, IplImage *dst)
{  
	IplImage *image_Re = 0, *image_Im = 0, *Fourier = 0; //实部、虚部


	image_Re = cvCreateImage(cvGetSize(src), IPL_DEPTH_64F, 1);  //实部


	image_Im = cvCreateImage(cvGetSize(src), IPL_DEPTH_64F, 1);  //虚部


	Fourier = cvCreateImage(cvGetSize(src), IPL_DEPTH_64F, 2);//2 channels (image_Re, image_Im)


	cvConvertScale(src, image_Re, 1, 0);// Real part conversion from u8 to 64f (double)


	cvZero(image_Im);// Imaginary part (zeros)


	cvMerge(image_Re, image_Im, 0, 0, Fourier);// Join real and imaginary parts and stock them in Fourier image


	cvDFT(Fourier, dst, CV_DXT_FORWARD);// Application of the forward Fourier transform


	cvReleaseImage(&image_Re);
	cvReleaseImage(&image_Im);
	cvReleaseImage(&Fourier);
}
void fft2shift(IplImage *src, IplImage *dst)
{
	IplImage *image_Re = 0, *image_Im = 0;
	int nRow, nCol, i, j, cy, cx;
	double scale, shift, tmp13, tmp24;


	image_Re = cvCreateImage(cvGetSize(src), IPL_DEPTH_64F, 1);


	image_Im = cvCreateImage(cvGetSize(src), IPL_DEPTH_64F, 1);
	cvSplit( src, image_Re, image_Im, 0, 0 );


	//具体原理见冈萨雷斯数字图像处理p123
	// Compute the magnitude of the spectrum Mag = sqrt(Re^2 + Im^2)
	//计算傅里叶谱
	cvPow( image_Re, image_Re, 2.0);
	cvPow( image_Im, image_Im, 2.0);
	cvAdd( image_Re, image_Im, image_Re);
	cvPow( image_Re, image_Re, 0.5 );


	//对数变换以增强灰度级细节(这种变换使以窄带低灰度输入图像值映射一宽带输出值，具体可见冈萨雷斯数字图像处理p62)
	// Compute log(1 + Mag);
	cvAddS( image_Re, cvScalar(1.0), image_Re ); // 1 + Mag
	cvLog( image_Re, image_Re ); // log(1 + Mag)


	//Rearrange the quadrants of Fourier image so that the origin is at the image center
	nRow = src->height; nCol = src->width;
	cx = nCol/2; cy = nRow/2; // image center


	//CV_IMAGE_ELEM为OpenCV定义的宏，用来读取图像的像素值，这一部分就是进行中心变换
	for( j = 0; j < cy; j++ ){
		for( i = 0; i < cx; i++ ){
			//中心化，将整体份成四块进行对角交换
			tmp13 = CV_IMAGE_ELEM( image_Re, double, j, i);
			CV_IMAGE_ELEM( image_Re, double, j, i) = CV_IMAGE_ELEM(image_Re, double, j+cy, i+cx);
			CV_IMAGE_ELEM( image_Re, double, j+cy, i+cx) = tmp13;


			tmp24 = CV_IMAGE_ELEM( image_Re, double, j, i+cx);
			CV_IMAGE_ELEM( image_Re, double, j, i+cx) =CV_IMAGE_ELEM( image_Re, double, j+cy, i);
			CV_IMAGE_ELEM( image_Re, double, j+cy, i) = tmp24;
		}
	}
	//归一化处理将矩阵的元素值归一为[0,255]
	//[(f(x,y)-minVal)/(maxVal-minVal)]*255
	double minVal = 0, maxVal = 0;
	// Localize minimum and maximum values
	cvMinMaxLoc( image_Re, &minVal, &maxVal );
	// Normalize image (0 - 255) to be observed as an u8 image
	scale = 255/(maxVal - minVal);
	shift = -minVal * scale;
	cvConvertScale(image_Re, dst, scale, shift);
	cvReleaseImage(&image_Re);
	cvReleaseImage(&image_Im);


}


/*
	for(int i=0;i<500;i++)
	{
		for(int j=0;j<500;j++)
		{
			if(i%20==0&&j%20==0)
				cout<<src1_f_re.at<float>(i,j)<<endl;
		}
	}
	*/

	/*string name1="lena.bmp";
	string name2="90du.bmp";
	cv::Mat src1=imread(name1);
	if(src1.channels()!=1)
		cv::cvtColor(src1,src1,CV_RGB2GRAY);
	cv::Mat src1_f_re   =cv::Mat::zeros(src1.size(),CV_32FC1);
	cv::Mat src1_f_im   =cv::Mat::zeros(src1.size(),CV_32FC1);
	cv::Mat src1_gonglv =cv::Mat::zeros(src1.size(),CV_32FC1);
	cv::Mat src1_g_log  =cv::Mat::zeros(src1.size(),CV_32FC1);
	fourier(src1,src1_f_re,src1_f_im);
	gonglvpu(src1_f_re,src1_f_im,src1_gonglv);
	cart2polar(src1_gonglv,src1_g_log);
	cv::normalize(src1_g_log,src1_g_log, 0, 1, CV_MINMAX);


	cv::Mat src2=imread(name2);
	if(src2.channels()!=1)
		cv::cvtColor(src2,src2,CV_RGB2GRAY);
	cv::Mat src2_f_re   =cv::Mat::zeros(src2.size(),CV_32FC1);
	cv::Mat src2_f_im   =cv::Mat::zeros(src2.size(),CV_32FC1);
	cv::Mat src2_gonglv =cv::Mat::zeros(src2.size(),CV_32FC1);
	cv::Mat src2_g_log  =cv::Mat::zeros(src2.size(),CV_32FC1);
	fourier(src2,src2_f_re,src2_f_im);
	gonglvpu(src2_f_re,src2_f_im,src2_gonglv);
	cart2polar(src2_gonglv,src2_g_log);
	cv::normalize(src2_g_log,src2_g_log, 0, 1, CV_MINMAX);
	int gx=0,gy=0;
	
	pingyi(src1_gonglv,src2_gonglv,gx,gy);
		//cout<<"max  "<<max<<"  min  "<<min<<endl;
	cout<<gx<<"   "<<gy<<endl;
	//pingyi(src1,src2,gx,gy);
	*/