码迷,mamicode.com
首页 > 其他好文 > 详细

特征提取代码总结

时间:2014-06-17 20:19:15      阅读:274      评论:0      收藏:0      [点我收藏+]

标签:des   class   code   http   tar   ext   

来自http://download.csdn.net/source/3208155#acomment

特征提取代码总结

颜色提取

Ø 颜色直方图提取:

Code:

 

#include <cv.h>

#include <highgui.h>

#include <iostream>

using namespace std;

 

 int main( int argc, char** argv )

{

IplImage * src= cvLoadImage("E:\\Download\\test1.jpg",1);

 

IplImage* hsv = cvCreateImage( cvGetSize(src), 8, 3 );

IplImage* h_plane = cvCreateImage( cvGetSize(src), 8, 1 );

IplImage* s_plane = cvCreateImage( cvGetSize(src), 8, 1 );

IplImage* v_plane = cvCreateImage( cvGetSize(src), 8, 1 );

IplImage* planes[] = { h_plane, s_plane };

 

/** H 分量划分为16个等级,S分量划分为8个等级*/

int h_bins = 16, s_bins = 8;

int hist_size[] = {h_bins, s_bins};

 

/** H 分量的变化范围*/

float h_ranges[] = { 0, 180 }; 

 

/** S 分量的变化范围*/

float s_ranges[] = { 0, 255 };

float* ranges[] = { h_ranges, s_ranges };

 

/** 输入图像转换到HSV颜色空间*/

cvCvtColor( src, hsv, CV_BGR2HSV );

cvCvtPixToPlane( hsv, h_plane, s_plane, v_plane, 0 );

 

/** 创建直方图,二维, 每个维度上均分*/

CvHistogram * hist = cvCreateHist( 2, hist_size, CV_HIST_ARRAY, ranges, 1 );

/** 根据H,S两个平面数据统计直方图*/

cvCalcHist( planes, hist, 0, 0 );

 

/** 获取直方图统计的最大值,用于动态显示直方图*/

float max_value;

cvGetMinMaxHistValue( hist, 0, &max_value, 0, 0 );

 

 

/** 设置直方图显示图像*/

int height = 240;

int width = (h_bins*s_bins*6);

IplImage* hist_img = cvCreateImage( cvSize(width,height), 8, 3 );

cvZero( hist_img );

 

/** 用来进行HSV到RGB颜色转换的临时单位图像*/

IplImage * hsv_color = cvCreateImage(cvSize(1,1),8,3);

IplImage * rgb_color = cvCreateImage(cvSize(1,1),8,3);

int bin_w = width / (h_bins * s_bins);

for(int h = 0; h < h_bins; h++)

{

for(int s = 0; s < s_bins; s++)

{

int i = h*s_bins + s;

/** 获得直方图中的统计次数,计算显示在图像中的高度*/

float bin_val = cvQueryHistValue_2D( hist, h, s );

int intensity = cvRound(bin_val*height/max_value);

 

/** 获得当前直方图代表的颜色,转换成RGB用于绘制*/

cvSet2D(hsv_color,0,0,cvScalar(h*180.f / h_bins,s*255.f/s_bins,255,0));

cvCvtColor(hsv_color,rgb_color,CV_HSV2BGR);

CvScalar color = cvGet2D(rgb_color,0,0);

 

cvRectangle( hist_img, cvPoint(i*bin_w,height),

cvPoint((i+1)*bin_w,height - intensity),

color, -1, 8, 0 );

}

}

 

cvNamedWindow( "Source", 1 );

cvShowImage( "Source", src );

 

cvNamedWindow( "H-S Histogram", 1 );

cvShowImage( "H-S Histogram", hist_img );

 

cvWaitKey(0);

}

运行效果截图:

                                                                             

 

形状提取

Ø Candy算子对边缘提取:

Code:

 

#include "cv.h"

#include "cxcore.h"

#include "highgui.h" 

int main( int argc, char** argv ) 

{

    //声明IplImage指针

         IplImage* pImg = NULL; 

    IplImage* pCannyImg = NULL; 

    //载入图像,强制转化为Gray

pImg = cvLoadImage( "E:\\Download\\test.jpg", 0); 

//为canny边缘图像申请空间

pCannyImg = cvCreateImage(cvGetSize(pImg), IPL_DEPTH_8U, 1);

//canny边缘检测

cvCanny(pImg, pCannyImg, 50, 150, 3); 

//创建窗口

cvNamedWindow("src", 1);

cvNamedWindow("canny",1);

//显示图像

cvShowImage( "src", pImg ); 

cvShowImage( "canny", pCannyImg ); 

//等待按键

cvWaitKey(0);

//销毁窗口

cvDestroyWindow( "src" ); 

cvDestroyWindow( "canny" );

//释放图像

cvReleaseImage( &pImg ); 

cvReleaseImage( &pCannyImg );

return 0;

 

运行效果截图:

 

 

 

Ø 角点提取:

Code:

 

#include <stdio.h>

#include "cv.h"

#include "highgui.h"

 

#define MAX_CORNERS 100

 

int main(void)

{

int cornersCount=MAX_CORNERS;//得到的角点数目

CvPoint2D32f corners[MAX_CORNERS];//输出角点集合

IplImage *srcImage = 0,*grayImage = 0,*corners1 = 0,*corners2 = 0;

int i;

CvScalar color = CV_RGB(255,0,0);

cvNamedWindow("image",1);

 

//Load the image to be processed

srcImage = cvLoadImage("E:\\Download\\1.jpg",1);

grayImage = cvCreateImage(cvGetSize(srcImage),IPL_DEPTH_8U,1);

 

//copy the source image to copy image after converting the format

//复制并转为灰度图像

cvCvtColor(srcImage,grayImage,CV_BGR2GRAY);

 

//create empty images os same size as the copied images

//两幅临时位浮点图像,cvGoodFeaturesToTrack会用到

corners1 = cvCreateImage(cvGetSize(srcImage),IPL_DEPTH_32F,1);

corners2 = cvCreateImage(cvGetSize(srcImage),IPL_DEPTH_32F,1);

 

cvGoodFeaturesToTrack(grayImage,corners1,corners2,corners,&cornersCount,0.05,

30,//角点的最小距离是

0,//整个图像

3,0,0.4);

printf("num corners found: %d\n",cornersCount);

 

//开始画出每个点

if (cornersCount>0)

{

for (i=0;i<cornersCount;i++)

{

cvCircle(srcImage,cvPoint((int)(corners[i].x),(int)(corners[i].y)),2,color,2,CV_AA,0);

}

}

cvShowImage("image",srcImage);

cvSaveImage("imagedst.png",srcImage);

 

cvReleaseImage(&srcImage);

cvReleaseImage(&grayImage);

cvReleaseImage(&corners1);

cvReleaseImage(&corners2);

 

cvWaitKey(0);

return 0;

运行效果截图:

 

 

 

Ø Hough直线提取:

Code:

 

#include <cv.h>

#include <highgui.h>

#include <math.h>

 

int main(int argc, char** argv)

{

    IplImage* src = cvLoadImage( "E:\\Download\\2.jpg" , 0 );

    IplImage* dst;

    IplImage* color_dst;

    CvMemStorage* storage = cvCreateMemStorage(0);

    CvSeq* lines = 0;

    int i;

 

    if( !src )

        return -1;

 

    dst = cvCreateImage( cvGetSize(src), 8, 1 );

    color_dst = cvCreateImage( cvGetSize(src), 8, 3 );

 

    cvCanny( src, dst, 50, 200, 3 );

    cvCvtColor( dst, color_dst, CV_GRAY2BGR );

#if 0

    lines = cvHoughLines2( dst, storage, CV_HOUGH_STANDARD, 1, CV_PI/180, 100, 0, 0 );

 

    for( i = 0; i < MIN(lines->total,100); i++ )

    {

        float* line = (float*)cvGetSeqElem(lines,i);

        float rho = line[0];

        float theta = line[1];

        CvPoint pt1, pt2;

        double a = cos(theta), b = sin(theta);

        double x0 = a*rho, y0 = b*rho;

        pt1.x = cvRound(x0 + 1000*(-b));

        pt1.y = cvRound(y0 + 1000*(a));

        pt2.x = cvRound(x0 - 1000*(-b));

        pt2.y = cvRound(y0 - 1000*(a));

        cvLine( color_dst, pt1, pt2, CV_RGB(255,0,0), 3, CV_AA, 0 );

    }

#else

    lines = cvHoughLines2( dst, storage, CV_HOUGH_PROBABILISTIC, 1, CV_PI/180, 50, 50, 10 );

    for( i = 0; i < lines->total; i++ )

    {

        CvPoint* line = (CvPoint*)cvGetSeqElem(lines,i);

        cvLine( color_dst, line[0], line[1], CV_RGB(255,0,0), 3, CV_AA, 0 );

    }

#endif

    cvNamedWindow( "Source", 1 );

    cvShowImage( "Source", src );

 

    cvNamedWindow( "Hough", 1 );

    cvShowImage( "Hough", color_dst );

 

    cvWaitKey(0);

 

    return 0;

}

运行效果截图:

 

 

 

Ø Hough圆提取:

Code:

 

#include <cv.h>

#include <highgui.h>

#include <math.h>

#include <iostream>

 

using namespace std;

 

int main(int argc, char** argv)

{

    IplImage* img;

img=cvLoadImage("E:\\Download\\3.jpg", 1);

 IplImage* gray = cvCreateImage( cvGetSize(img), 8, 1 );

     CvMemStorage* storage = cvCreateMemStorage(0);

     cvCvtColor( img, gray, CV_BGR2GRAY );

     cvSmooth( gray, gray, CV_GAUSSIAN, 5, 15 );

// smooth it, otherwise a lot of false circles may be detected

CvSeq* circles = cvHoughCircles( gray, storage, CV_HOUGH_GRADIENT, 2, gray->height/4, 200, 100 );

    int i;

     for( i = 0; i < circles->total; i++ )

     {

          float* p = (float*)cvGetSeqElem( circles, i );

          cvCircle( img, cvPoint(cvRound(p[0]),cvRound(p[1])), 3, CV_RGB(0,255,0), -1, 8, 0 );

 cvCircle( img, cvPoint(cvRound(p[0]),cvRound(p[1])), cvRound(p[2]), CV_RGB(255,0,0), 3, 8, 0 );

          cout<<"圆心坐标x= "<<cvRound(p[0])<<endl<<"圆心坐标y= "<<cvRound(p[1])<<endl;

          cout<<"半径="<<cvRound(p[2])<<endl; 

     }

     cout<<"圆数量="<<circles->total<<endl;

     cvNamedWindow( "circles", 1 );

     cvShowImage( "circles", img );

     cvWaitKey(0);

  

    return 0;

}

 

运行效果截图:

 

 

 

Ø Hough矩形提取:

Code:

 

#include "cv.h"

#include "highgui.h"

#include <stdio.h>

#include <math.h>

#include <string.h>

 

int thresh = 50;

IplImage* img = 0;

IplImage* img0 = 0;

CvMemStorage* storage = 0;

CvPoint pt[4];const char* wndname = "Square Detection Demo"; 

 

double angle( CvPoint* pt1, CvPoint* pt2, CvPoint* pt0 )

{    

double dx1 = pt1->x - pt0->x; 

double dy1 = pt1->y - pt0->y;  

double dx2 = pt2->x - pt0->x;  

double dy2 = pt2->y - pt0->y;    

return (dx1*dx2 + dy1*dy2)/sqrt((dx1*dx1 + dy1*dy1)*(dx2*dx2 + dy2*dy2) + 1e-10);

CvSeq* findSquares4( IplImage* img, CvMemStorage* storage )

{  

CvSeq* contours;

int i, c, l, N = 11;  

CvSize sz = cvSize( img->width & -2, img->height & -2 );

IplImage* timg = cvCloneImage( img );

IplImage* gray = cvCreateImage( sz, 8, 1 ); 

IplImage* pyr = cvCreateImage( cvSize(sz.width/2, sz.height/2), 8, 3 );  

IplImage* tgray;   

CvSeq* result;  

double s, t;  

CvSeq* squares = cvCreateSeq( 0, sizeof(CvSeq), sizeof(CvPoint), storage );   

cvSetImageROI( timg, cvRect( 0, 0, sz.width, sz.height ));   

// down-scale and upscale the image to filter out the noise 

cvPyrDown( timg, pyr, 7 );  

cvPyrUp( pyr, timg, 7 );   

tgray = cvCreateImage( sz, 8, 1 );  

// find squares in every color plane of the image 

for( c = 0; c < 3; c++ )  

{       

cvSetImageCOI( timg, c+1 );     

cvCopy( timg, tgray, 0 );           

for( l = 0; l < N; l++ )     

{          

if( l == 0 )     

{               

cvCanny( tgray, gray, 0, thresh, 5 );        

cvDilate( gray, gray, 0, 1 );      

}           

else       

{             

cvThreshold( tgray, gray, (l+1)*255/N, 255, CV_THRESH_BINARY );

}                       

cvFindContours( gray, storage, &contours, sizeof(CvContour),CV_RETR_LIST, CV_CHAIN_APPROX_SIMPLE, cvPoint(0,0) );          

while( contours )    

{              

  result = cvApproxPoly( contours, sizeof(CvContour), storage,CV_POLY_APPROX_DP, cvContourPerimeter(contours)*0.02, 0 ); 

if( result->total == 4 && fabs(cvContourArea(result,CV_WHOLE_SEQ)) > 1000 &&  cvCheckContourConvexity(result) )  

{               

s = 0;      

for( i = 0; i < 5; i++ )  

{                   

if( i >= 2 )           

{               

t = fabs(angle( (CvPoint*)cvGetSeqElem( result, i ),(CvPoint*)cvGetSeqElem( result, i-2 ),(CvPoint*)cvGetSeqElem( result, i-1 )));   

s = s > t ? s : t;     

}         

}                                            

if( s < 0.3 )                      

for( i = 0; i < 4; i++ )              

cvSeqPush( squares,                    

(CvPoint*)cvGetSeqElem( result, i ));     

}                                      

contours = contours->h_next;      

}   

}

cvReleaseImage( &gray );   

cvReleaseImage( &pyr );  

cvReleaseImage( &tgray );  

cvReleaseImage( &timg );   

return squares;

}  

// the function draws all the squares in the image

void drawSquares( IplImage* img, CvSeq* squares )

{   

CvSeqReader reader;   

IplImage* cpy = cvCloneImage( img );   

int i;       

cvStartReadSeq( squares, &reader, 0 );     

for( i = 0; i < squares->total; i += 4 )  

{       

CvPoint* rect = pt;    

int count = 4;      

memcpy( pt, reader.ptr, squares->elem_size ); 

CV_NEXT_SEQ_ELEM( squares->elem_size, reader ); 

memcpy( pt + 1, reader.ptr, squares->elem_size );     

CV_NEXT_SEQ_ELEM( squares->elem_size, reader );   

memcpy( pt + 2, reader.ptr, squares->elem_size );   

CV_NEXT_SEQ_ELEM( squares->elem_size, reader );     

memcpy( pt + 3, reader.ptr, squares->elem_size );  

CV_NEXT_SEQ_ELEM( squares->elem_size, reader );         

cvPolyLine( cpy, &rect, &count, 1, 1, CV_RGB(0,255,0), 3, CV_AA, 0 );

}        

cvShowImage( wndname, cpy );  

cvReleaseImage( &cpy );

}

void on_trackbar( int a )

{   

if( img )    

drawSquares( img, findSquares4( img, storage ) );

}

char* names[] = { "1.jpg", 0 };

int main(int argc, char** argv)

{    

int i, c; 

storage = cvCreateMemStorage(0);    

for( i = 0; names[i] != 0; i++ )   

{     

img0 = cvLoadImage( names[i], 1 );   

if( !img0 )    

{        

printf("Couldn‘t load %s\n", names[i] );    

continue;     

}     

img = cvCloneImage( img0 );       

     cvNamedWindow( wndname, 1 );     

cvCreateTrackbar( "canny thresh", wndname, &thresh, 1000, on_trackbar );     

on_trackbar(0);       

c = cvWaitKey(0);     

cvReleaseImage( &img );    

cvReleaseImage( &img0 );       

cvClearMemStorage( storage );      

if( c == 27 )       

break;   

}       

cvDestroyWindow( wndname );   

return 0;

运行效果截图:

 

 

 

Ø 边缘直方图提取:

Code:

  

#include "cv.h"

#include "highgui.h"

#include <stdio.h>

#include <ctype.h>

#define PI 3.14

 

int main()

{

    IplImage *src = 0;  // source imagre

    IplImage *histimg = 0; // histogram image 

    CvHistogram *hist = 0; // define multi_demention histogram

    IplImage* canny;

    CvMat* canny_m;

    IplImage* dx; // the sobel x difference 

    IplImage* dy; // the sobel y difference 

    CvMat* gradient; // value of gradient

    CvMat* gradient_dir; // direction of gradient

    CvMat* dx_m; // format transform to matrix

    CvMat* dy_m;

    CvMat* mask;

    CvSize  size;

    IplImage* gradient_im;

    int i,j;

    float theta;

    

    int hdims = 8;     // 划分HIST的个数,越高越精确

    float hranges_arr[] = {-PI/2,PI/2}; // 直方图的上界和下界

    float* hranges = hranges_arr;

                                                                                                                                                                                                                                                              

    float max_val;  // 

    int bin_w;

    

    src=cvLoadImage("E:\\Download\\test.jpg", 0); // force to gray image

       if(src==0) return -1;

    

    cvNamedWindow( "Histogram", 0 );

    //cvNamedWindow( "src", 0);

    size=cvGetSize(src);

    canny=cvCreateImage(cvGetSize(src),8,1);//边缘图像

    dx=cvCreateImage(cvGetSize(src),32,1);//x方向上的差分//此处的数据类型为U 不怕溢出吗?

    dy=cvCreateImage(cvGetSize(src),32,1);

    gradient_im=cvCreateImage(cvGetSize(src),32,1);//梯度图像

    canny_m=cvCreateMat(size.height,size.width,CV_32FC1);//边缘矩阵

    dx_m=cvCreateMat(size.height,size.width,CV_32FC1);

    dy_m=cvCreateMat(size.height,size.width,CV_32FC1);

    gradient=cvCreateMat(size.height,size.width,CV_32FC1);//梯度矩阵

    gradient_dir=cvCreateMat(size.height,size.width,CV_32FC1);//梯度方向矩阵

    mask=cvCreateMat(size.height,size.width,CV_32FC1);//掩码

 

    cvCanny(src,canny,60,180,3);//边缘检测

    cvConvert(canny,canny_m);//把图像转换为矩阵

    cvSobel(src,dx,1,0,3);// 一阶X方向的图像差分:dx

    cvSobel(src,dy,0,1,3);// 一阶Y方向的图像差分:dy

    cvConvert(dx,dx_m);

    cvConvert(dy,dy_m);

    cvAdd(dx_m,dy_m,gradient); // value of gradient//梯度不是等于根号下x的导数的平方加上y导数的平方吗?

    cvDiv(dx_m,dy_m,gradient_dir); // direction

    for(i=0;i<size.height;i++)

    for(j=0;j<size.width;j++)

    {

      if(cvmGet(canny_m,i,j)!=0 && cvmGet(dx_m,i,j)!=0)//此行是什么意思?只看边缘上的方向?

      {

         theta=cvmGet(gradient_dir,i,j);

         theta=atan(theta);

         cvmSet(gradient_dir,i,j,theta);  

      }

      else

      {

         cvmSet(gradient_dir,i,j,0);

      }

         

    }

   hist = cvCreateHist( 1, &hdims, CV_HIST_ARRAY, &hranges, 1 );  

// 创建一个指定尺寸的直方图,并返回创建的直方图指针

   histimg = cvCreateImage( cvSize(320,200), 8, 3 ); // 创建一个图像,通道

   cvZero( histimg ); // 清;

   cvConvert(gradient_dir,gradient_im);//把梯度方向矩阵转化为图像

   cvCalcHist( &gradient_im, hist, 0, canny ); // 计算直方图

   cvGetMinMaxHistValue( hist, 0, &max_val, 0, 0 );  // 只找最大值

   cvConvertScale( hist->bins, hist->bins, max_val ? 255. / max_val : 0., 0 ); 

// 缩放bin 到区间[0,255] ,比例系数

   cvZero( histimg );

   bin_w = histimg->width /16;  // hdims: 条的个数,则bin_w 为条的宽度

    

    // 画直方图

    for( i = 0; i < hdims; i++ )

    {

       double val = ( cvGetReal1D(hist->bins,i)*histimg->height/255 );

// 返回单通道数组的指定元素, 返回直方图第i条的大小,val为histimg中的i条的高度

        CvScalar color = CV_RGB(255,255,0); //(hsv2rgb(i*180.f/hdims);//直方图颜色

        cvRectangle( histimg, cvPoint(100+i*bin_w,histimg->height),cvPoint(100+(i+1)*bin_w,(int)(histimg->height - val)), color, 1, 8, 0 ); // 画直方图——画矩形,左下角,右上角坐标

     }

    

    cvShowImage( "src", src);

    cvShowImage( "Histogram", histimg );

    cvWaitKey(0);

 

    cvDestroyWindow("src");

    cvDestroyWindow("Histogram");

    cvReleaseImage( &src );

    cvReleaseImage( &histimg );

    cvReleaseHist ( &hist );

    

    return 0;

}

运行效果截图:

 

 

 

Ø 视频流中边缘检测:

Code:

 

#include "highgui.h"

#include "cv.h"

#include "stdio.h"

#include <ctype.h> 

int main(int argc,char ** argv)

{

       IplImage * laplace = 0;

       IplImage * colorlaplace = 0;

       IplImage * planes[3] = {0,0,0};

 

       CvCapture *capture = 0;

 

      //从摄像头读取

       /*if(argc == 1 ||( argc==2 && strlen(argv[1])==1 && isdigit(argv[1][0]) ))

           capture = cvCaptureFromCAM(argc == 2 ? argv[1][0] -‘0‘:0);*/

       //从文件中读取

      /* else if(argc == 2)*/

           capture = cvCaptureFromAVI("1.avi");

       if(!capture)

       {

           fprintf(stderr,"Could not initialize capturing...\n");

           return -1;

       }

       cvNamedWindow("Laplacian",1);

       cvNamedWindow("video",1);

       //循环捕捉,直到用户按键跳出循环体

       for(;;)

       {

            IplImage * frame =0;    //抓起一祯

            frame = cvQueryFrame(capture);

            if(!frame)

                 break;

            if(!laplace)

            {

               //创建图像

               for(int i=0;i<3;i++)

               planes[i] = cvCreateImage(cvSize(frame->width,frame->height),IPL_DEPTH_8U,1);

               laplace = cvCreateImage(cvSize(frame->width,frame->height),IPL_DEPTH_16S,1);

             colorlaplace=cvCreateImage(cvSize(frame->width,frame->height),IPL_DEPTH_8U,3);

            }

             cvCvtPixToPlane(frame,planes[0],planes[1],planes[2],0);

             for(int i=0;i<3;i++)

             {

                //交换,如通道变换

                cvLaplace(planes[i],laplace,3);

                //使用线性变换转换输入函数元素成为无符号整形

                cvConvertScaleAbs(laplace,planes[i],1,0);

              }

            cvCvtPlaneToPix(planes[0],planes[1],planes[2],0,colorlaplace);

            //结构相同(- 顶—左结构,1 - 底—左结构)

            colorlaplace->origin = frame->origin;

           //高斯滤波,平滑图像

           // cvSmooth(colorlaplace, colorlaplace, CV_GAUSSIAN, 1, 0, 0);

           //形态学滤波,闭运算

           cvDilate(colorlaplace, colorlaplace, 0, 1);//膨胀

           cvErode(colorlaplace, colorlaplace, 0, 1);//腐蚀

           cvShowImage("video", frame);

           cvShowImage("Laplacian",colorlaplace);

           if(cvWaitKey(10)>0)

               break;

       }

       cvReleaseCapture(&capture);

       cvDestroyWindow("Laplacian");

       cvDestroyWindow("video");

       return 0;

}

运行效果截图:

 

 

 

Ø 纹理提取:

Code:

 

#include <iostream>

#include <math.h>

#include "cv.h"

#include "highgui.h"

 

int main(int argc, char* argv[])

{

int tmp[8]={0};

int sum=0;int k=0;

IplImage* img,*dst;

img=cvLoadImage("E:\\Download\\2.jpg",0);

CvScalar s;

cvNamedWindow("img",NULL);

cvNamedWindow("dst",NULL);

cvShowImage("img",img);

 

uchar* data=(uchar*)img->imageData;

int step=img->widthStep;

dst=cvCreateImage(cvSize(img->width,img->height),img->depth,1);

dst->widthStep=img->widthStep;

for(int i=1;i<img->height-1;i++)

for(int j=1;j<img->width-1;j++)

{

if(data[(i-1)*step+j-1]>data[i*step+j]) tmp[0]=1;

else tmp[0]=0;

if(data[i*step+(j-1)]>data[i*step+j]) tmp[1]=1;

else tmp[1]=0;

if(data[(i+1)*step+(j-1)]>data[i*step+j]) tmp[2]=1;

else tmp[2]=0;

if (data[(i+1)*step+j]>data[i*step+j]) tmp[3]=1;

else tmp[3]=0;

if (data[(i+1)*step+(j+1)]>data[i*step+j]) tmp[4]=1;

else tmp[4]=0;

if(data[i*step+(j+1)]>data[i*step+j]) tmp[5]=1;

else tmp[5]=0;

if(data[(i-1)*step+(j+1)]>data[i*step+j]) tmp[6]=1;

else tmp[6]=0;

if(data[(i-1)*step+j]>data[i*step+j]) tmp[7]=1;

else tmp[7]=0;

for(k=0;k<=7;k++)

sum+=abs(tmp[k]-tmp[k+1]);

sum=sum+abs(tmp[7]-tmp[0]);

if (sum<=2)

s.val[0]=(tmp[0]*128+tmp[1]*64+tmp[2]*32+tmp[3]*16+tmp[4]*8+tmp[5]*4+tmp[6]*2+tmp[7]);

else s.val[0]=59; 

cvSet2D(dst,i,j,s);

}

 

cvShowImage("dst",dst);

cvWaitKey(-1);

 

return 0;

}

运行效果截图:

 

特征提取代码总结,布布扣,bubuko.com

特征提取代码总结

标签:des   class   code   http   tar   ext   

原文地址:http://www.cnblogs.com/gune/p/3792178.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!