HOG特征提取

#include<cv.h> #include<highgui.h> #include<math.h> #include<opencv.hpp> using namespace cv; using namespace std; IplImage * GRBtoGrey(IplImage* img)//RGB图像转为灰度图像 { IplImage* dest = cvCreateImage(cvSize(img->width, img->height), img->depth, 1); uchar* data = (uchar*)img->imageData; int b, g, r; for (int i = 0; i<img->height; i++){ for (int j = 0; j<img->width; j++){ b = data[i*img->widthStep + j*img->nChannels + 0]; g = data[i*img->widthStep + j*img->nChannels + 1]; r = data[i*img->widthStep + j*img->nChannels + 2]; ((uchar*)(dest->imageData + i*dest->widthStep))[j] = 0.3*r + 0.59*g + 0.11*b; } } return dest; } IplImage* nomolize(IplImage *img)//必须是灰度图 { uchar* gM = (uchar*)img->imageData; int max = 0, min = 255; for (int i = 0; i<img->height; i++){ for (int j = 0; j<img->width; j++){ int num = gM[i*img->widthStep + j*img->nChannels]; if (num>max){ max = num; } if (num<min){ min = num; } } } for (int i = 0; i<img->height; i++){ for (int j = 0; j<img->width; j++){ int num = (int)(float)(gM[i*img->widthStep + j*img->nChannels] - min)* ((float)(255) / (float)(max - min)); if (num>255){ num = 255; } ((uchar*)(img->imageData + i*img->widthStep))[j] = num; } } return img; } IplImage *greygamma(IplImage *greyimage){ IplImage *gammaimage = cvCreateImage(cvSize(greyimage->width, greyimage->height), greyimage->depth, greyimage->nChannels); uchar* data = (uchar*)greyimage->imageData; for (int i = 0; i<greyimage->height; i++){ for (int j = 0; j<greyimage->width; j++){ int gamma = data[i*greyimage->widthStep + j*greyimage->nChannels]; ((uchar*)(gammaimage->imageData + i*gammaimage->widthStep))[j] = (int)sqrt((float)gamma); } } return nomolize(gammaimage); } IplImage *gradientimg(IplImage *img){ uchar * data = (uchar*)img->imageData; for (int i = 1; i<img->height - 1; i++){ for (int j = 1; j<img->width - 1; j++){ int Gx = data[i*img->widthStep + (j + 1)*img->nChannels] - data[i*img->widthStep + (j - 1)*img->nChannels]; int Gy = data[(i + 1)*img->widthStep + j*img->nChannels] - data[(i - 1)*img->widthStep + j*img->nChannels]; int gradient = (int)sqrt((float)(Gx*Gx + Gy*Gy)); int theta = (int)(atan((float)Gy / (float)Gx) * 180 / 3.14); ((uchar*)(img->imageData + i*img->widthStep))[j] = theta; } } return nomolize(img); } int *CellFeature(IplImage *img, int w, int h){//8*8 cell int cell_w = 8; int f_n = 9; int *feature = new int[f_n]; for (int i = 0; i<f_n; i++){ f_n = 0; } for (int i = h; i<h + cell_w; i++){ for (int j = w; j<w + cell_w; j++){ feature[((uchar*)(img->imageData + i*img->widthStep))[j] / 40]++; } } return feature; } int *HogFeature(IplImage *img){ int cell_w = 8; int cell_fn = 9; int cell_wn = img->width / cell_w; int cell_hn = img->height / cell_w; int *hog_f = new int[cell_wn*cell_hn*cell_fn]; int m = 0; for (int i = 0; i<cell_hn; i++){ for (int j = 0; j<cell_wn; j++){ int *feature = CellFeature(img, j*cell_w, i); for (int k = 0; k<cell_fn; k++){ hog_f[m++] = feature[k]; } } } for (int i = 0; i<cell_wn*cell_hn*cell_fn; i++){ if (i % 9 == 0){ printf("\n"); } if (i % (2 * 2 * 9) == 0){ printf("-------------------------------------\n\n"); } printf("%d ", hog_f[i]); } return hog_f; } int main() { IplImage *img = cvLoadImage("Img2.jpg"); if (img == NULL){ printf("img is null"); return 0; } cvNamedWindow("greyimage"); IplImage* greyimg = GRBtoGrey(img); cvShowImage("greyimage",greyimg); cvWaitKey(); cvNamedWindow("gammaimage"); IplImage* gammaimg = greygamma(greyimg); cvShowImage("gammaimage",gammaimg); printf("gamma end\n"); cvWaitKey(); cvNamedWindow("gradientimage"); IplImage* gradientimage = gradientimg(gammaimg); cvShowImage("gradientimage",gradientimage); printf("gradient end\n"); cvWaitKey(); int *hog_f = HogFeature(gradientimage); system("pause"); return 1; }