立体匹配之NCC算法

FROM：http://blog.csdn.net/tulun/article/details/6388759

NCC算法（Normal Cross Correlation），具体原理见相关图像处理书籍。

该程序是opencv中文论坛的牛人贡献的，感谢他的工作。

（程序所需图片可以在网上找如http://vision.middlebury.edu/stereo/data/scenes2003/）

#include <iostream>
#include <stdio.h>
#include <stdlib.h>
#include <cv.h>
#include <cxcore.h>
#include <highgui.h>
#include <math.h>
#include <ctime>

using namespace std;
template<class T> class Image
{
private:
IplImage* imgp;

public:
    Image(IplImage* img=0){imgp=img;}
    ~Image(){imgp=0;}
    void operator=(IplImage* img){imgp=img;}
    inline T* operator[](const int rowIndx)
    {
        return((T*)(imgp->imageData+rowIndx*imgp->widthStep));
    }

};

typedef struct
{
unsigned char b,g,r;
}RgbPixel;

typedef struct
{
float b,g,r;
}RgbPixelFloat;

typedef Image<RgbPixel> RgbImage;
typedef Image<RgbPixelFloat> RgbImageFloat;
typedef Image<unsigned char> BwImage;
typedef Image<float> BwImageFloat;

void displayImageProperty(IplImage* image)
{
    cout<<"-------Image Properties--------"<<endl;
    cout<<"Image width="<<image->width<<endl;
    cout<<"Image height="<<image->height<<endl;
    cout<<"Image depth="<<image->depth<<endl;
    cout<<"Image nSize="<<image->nSize<<endl;
    cout<<"Image nChannels="<<image->nChannels<<endl;

char* origin;
    char* dataOrder;
    if (image->origin==0)
    {
        origin="Top-left";
    }
    else
    {
        origin="Below-left";//image->origin=1
    }

cout<<"Image origin="<<origin<<endl;

if (image->dataOrder==0)
    {
        dataOrder="Order_Pixel(Interleaved)";
    }
    else
    {
        dataOrder="Order_Plane";//image->dataOrder=1
    }
    cout<<"Image dataOrder="<<dataOrder<<endl;
    cout<<"Image widthStep="<<image->widthStep<<" Bytes"<<endl;
}

// display an image in a new window with title to be given.
void displayImageNewWindow(char* title,CvArr* img)
{
cvNamedWindow(title, CV_WINDOW_AUTOSIZE );
cvShowImage(title,img);
}

int getMaxMin(double value[],int valueSize, int maxmin)
{
    int pos=0;
    int i=0;
    double max1=-1;//?-999999;
    double min1=999999;

    if (maxmin==1)
    {
        //find max
        for (i=0;i<valueSize;i++)
        {
            //find the index with the max ncc;
            if (value[i]>max1)
            {
                pos=i;
                max1=value[i];
            }
         }
    }

    if (maxmin==0)
    {
        //find min
        for (i=0;i<valueSize;i++)
        {
            //find the index with the max ncc;
            if (value[i]<min1)
            {
                pos=i;
                min1=value[i];
            }
         }
    }
    return pos;
}

IplImage* generateDisparityImage(IplImage* greyLeftImg32,IplImage* greyRightImg32,int windowSize,int DSR)
{
    int offset=floor((double)windowSize/2);
    int height=greyLeftImg32->height;
    int width=greyLeftImg32->width;
    double* localNCC=new double[DSR];

int x=0, y=0,d=0,m=0;
int N=windowSize;

IplImage* leftWinImg=cvCreateImage(cvSize(N,N),32,1);//mySubImage(greyLeftImg32,cvRect(0,0,N,N));
    IplImage* rightWinImg=cvCreateImage(cvSize(N,N),32,1);;//mySubImage(greyRightImg32,cvRect(0,0,N,N));
    IplImage* disparity=cvCreateImage(cvSize(width,height),8,1);//or IPL_DEPTH_8U
    BwImage imgA(disparity);

    for (y=0;y<height;y++)
    {
        for (x=0;x<width;x++)
        {
            imgA[y][x]=0;
        }
    }

    CvScalar s1;
    CvScalar s2;
    for (y=0;y<height-N;y++)
    {
        //height-N
        for (x=0;x<width-N;x++)
        {
            //width-N
            //getWindow(i,j,leftim,wl,N);
            cvSetImageROI(greyLeftImg32, cvRect(x,y,N,N));
            s1=cvAvg(greyLeftImg32,NULL);
            cvSubS(greyLeftImg32,s1,leftWinImg,NULL);//zero-means
            cvNormalize(leftWinImg,leftWinImg,1,0,CV_L2,NULL);//0变成1
            d=0;

            //initialise localNCC
            for (m=0;m<DSR;m++)
            {
                localNCC[m]=0;
            }

            do{
                if (x-d>=0)
                {
                    cvSetImageROI(greyRightImg32, cvRect(x-d,y,N,N));
                    s2=cvAvg(greyRightImg32,NULL);
                    cvSubS(greyRightImg32,s2,rightWinImg,NULL);//zero-means
                    cvNormalize(rightWinImg,rightWinImg,1,0,CV_L2,NULL);//0变成1
                }
                else
                {
                    break;
                }
                localNCC[d]=cvDotProduct(leftWinImg,rightWinImg);
                cvResetImageROI(greyRightImg32);
                d++;
             }while(d<=DSR);

             //to find the best d and store
             imgA[y+offset][x+offset]=getMaxMin(localNCC,DSR,1)*16;
             cvResetImageROI(greyLeftImg32);
          }//x
          if (y%10==0)
            cout<<"row="<<y<<" of "<<height<<endl;
       }//y

       cvReleaseImage(&leftWinImg);
       cvReleaseImage(&rightWinImg);

       return disparity;
}

int main (int argc, char * const argv[])
{
    // insert code here...
    cout << "Stereo Normalized Cross Correlation"<<endl;

    //**********image input*********************//

char* filename1="im0.ppm";//im2_cone.png
    IplImage* greyLeftImg= cvLoadImage(filename1,0);
    char* filename2="im1.ppm";
    IplImage* greyRightImg= cvLoadImage(filename2,0);

    if (greyLeftImg==NULL)
    {
        cout << "No valid image input."<<endl;
        //char c=getchar();
        return 1;
    }
    else
    {
        displayImageProperty(greyLeftImg);
    }

    if (greyRightImg==NULL)
    {
        cout << "No valid image input."<<endl;
        //char c=getchar();
        return 1;
    }

int width=greyLeftImg->width;
int height=greyLeftImg->height;

/****************8U to 32F**********************/
    IplImage* greyLeftImg32=cvCreateImage(cvSize(width,height),32,1);//IPL_DEPTH_32F
    IplImage* greyRightImg32=cvCreateImage(cvSize(width,height),32,1);
    cvConvertScale(greyLeftImg, greyLeftImg32, 1/255.);
    cvConvertScale(greyRightImg, greyRightImg32, 1/255.);//1/255. equals to 1/255.0

    //-------------Computing stereo matching----------------
    time_t tstart, tend;
    tstart = time(0);
    int windowSize=11,DSR=20;//Disparity Search Range
    IplImage* disparity32=generateDisparityImage(greyLeftImg32,greyRightImg32,windowSize,DSR);
    tend = time(0);
    cout << "It took "<< difftime(tend, tstart) <<" second(s)."<< endl;

displayImageNewWindow("Dispairty Image",disparity32);
    displayImageNewWindow("Left Image",greyLeftImg32);
    displayImageNewWindow("Right Image",greyRightImg32);
    //cvSaveImage("D:/OpenCV_stuff/SampleImages/disparity.jpg",disparity32);

//********destroy window************/
    cvWaitKey(0);
    cvReleaseImage(&greyLeftImg32);
    cvReleaseImage(&greyRightImg32);
    cvReleaseImage(&greyLeftImg);
    cvReleaseImage(&greyRightImg);
    cvReleaseImage(&disparity32);
    cvDestroyWindow("Left Image");
    cvDestroyWindow("Right Image");
    cvDestroyWindow("Dispairty Image");
    return 0;
}

立体匹配之NCC算法相关推荐

python计算机视觉——立体匹配与NCC算法
文章目录 1. 立体匹配 1.1 概述 1.2 主要立体匹配算法分类 1.3 立体匹配的基本步骤 2. 归一化互相关(NCC)视差匹配法 2.1 原理 2.2 匹配流程 2.3 代码实现 3. 不同窗 ...
图像匹配—NCC算法，即归一化互相关匹配
一.NCC的基础概念 NCC(normalized cross correlation)算法,归一化互相关匹配法,是基于图像灰度信息的匹配方法. 二.NCC算法定义为: 三.代码(利用NCC的密集匹配 ...
立体匹配：经典算法Fast Bilateral Solver
点击上方"3D视觉工坊",选择"星标" 干货第一时间送达作者丨HawkWang 来源丨计算摄影学点击进入->3D视觉工坊学习交流群一. 前言你好, ...
图像匹配算法 MAD、SAD、SSD、MSD、NCC、SSDA、SATD,LBD算法
图像匹配算法分为3类:基于灰度的匹配算法.基于特征的匹配算法.基于关系的匹配算法 (1)基于灰度的模板匹配算法:模板匹配(Blocking Matching)是根据已知模板图像到另一幅图像中寻找与模板 ...
基于灰度的模板匹配算法（一）：MAD、SAD、SSD、MSD、NCC、SSDA算法
简介: 本文主要介绍几种基于灰度的图像匹配算法:平均绝对差算法(MAD).绝对误差和算法(SAD).误差平方和算法(SSD).平均误差平方和算法(MSD).归一化积相关算法(NCC).序贯相似性算法( ...
【图像配准】基于灰度的模板匹配算法（一）：MAD、SAD、SSD、MSD、NCC、SSDA、SATD算法
简介: 本文主要介绍几种基于灰度的图像匹配算法:平均绝对差算法(MAD).绝对误差和算法(SAD).误差平方和算法(SSD).平均误差平方和算法(MSD).归一化积相关算法(NCC).序贯相似性检测算 ...
基于灰度的模板匹配算法：MAD、SAD、SSD、MSD、NCC、SSDA算法
简介: 本文主要介绍几种基于灰度的图像匹配算法:平均绝对差算法(MAD).绝对误差和算法(SAD).误差平方和算法(SSD).平均误差平方和算法(MSD).归一化积相关算法(NCC).序贯相似性算法( ...
【计算机视觉】NCC匹配算法
NCC匹配算法 1.NCC匹配实验原理 1.1 NCC的基础概念 1.2 NCC算法的基本原理 1.3 相关的数学知识 1.4 双目立体匹配流程 1.4 NCC的特点 2.NCC算法实现视差图匹配实验 ...
【视觉项目】【day6】8.26关于matchTemplate()以及NCC的思考整理
NCC与matchTemplate()函数中match_method = TM_CCOEFF_NORMED是否一样? 先看公式: TM_CCOEFF_NORMED NCC TM_CCOEFF_NORM ...

立体匹配之NCC算法

立体匹配之NCC算法相关推荐

最新文章

热门文章