如何计算2张图像中相同元素之间的像素距离
问题描述:
我想用C编写程序以确定两张图像之间的平移所拍摄的两张图像中相同元素之间的距离(通过立体图测量距离) 。如何计算2张图像中相同元素之间的像素距离
我正在用逐像素计算的两幅图像之间的估计“距离”计算该距离。但是,这真的很慢。
我听说过使用称为交叉相关和FFT的方法能更快地完成它的方法。但我无法在网上找到任何代码或信息。
你有一些信息吗?
谢谢!
P.S. :我使用OpenCV加载和处理图像。
答
感谢您的回答,我设法编写了一个几乎可以工作的代码。
struct Peak Find_FFT(IplImage* src, IplImage* tpl, int hamming)
{
int i, j, k;
double tmp; //To store the modulus temporarily
//src and tpl must be the same size
assert(src->width == tpl->width);
assert(src->height == tpl->height);
// Get image properties
int width = src->width;
int height = src->height;
int step = src->widthStep;
int fft_size = width * height;
//fftw_init_threads(); //Initialize FFTW for multithreading with a max number of 4 threads
//fftw_plan_with_nthreads(4);
//Allocate arrays for FFT of src and tpl
fftw_complex *src_spatial = (fftw_complex*)fftw_malloc(sizeof(fftw_complex) * width * height);
fftw_complex *src_freq = (fftw_complex*)fftw_malloc(sizeof(fftw_complex) * width * height);
fftw_complex *tpl_spatial = (fftw_complex*)fftw_malloc(sizeof(fftw_complex) * width * height);
fftw_complex *tpl_freq = (fftw_complex*)fftw_malloc(sizeof(fftw_complex) * width * height);
fftw_complex *res_spatial = (fftw_complex*)fftw_malloc(sizeof(fftw_complex) * width * height); //Result = Cross correlation
fftw_complex *res_freq = (fftw_complex*)fftw_malloc(sizeof(fftw_complex) * width * height);
// Setup pointers to images
uchar *src_data = (uchar*) src->imageData;
uchar *tpl_data = (uchar*) tpl->imageData;
// Fill the structure that will be used by fftw
for(i = 0; i < height; i++)
{
for(j = 0 ; j < width ; j++, k++)
{
src_spatial[k][0] = (double) src_data[i * step + j];
src_spatial[k][1] = 0.0;
tpl_spatial[k][0] = (double) tpl_data[i * step + j];
tpl_spatial[k][1] = 0.0;
}
}
// Hamming window to improve FFT (but slightly slower to compute)
if(hamming == 1)
{
double omega = 2.0*M_PI/(fft_size-1);
double A= 0.54;
double B= 0.46;
for(i=0,k=0;i<height;i++)
{
for(j=0;j<width;j++,k++)
{
src_spatial[k][0]= (src_spatial[k][0])*(A-B*cos(omega*k));
tpl_spatial[k][0]= (tpl_spatial[k][0])*(A-B*cos(omega*k));
}
}
}
// Setup FFTW plans
fftw_plan plan_src = fftw_plan_dft_2d(height, width, src_spatial, src_freq, FFTW_FORWARD, FFTW_ESTIMATE);
fftw_plan plan_tpl = fftw_plan_dft_2d(height, width, tpl_spatial, tpl_freq, FFTW_FORWARD, FFTW_ESTIMATE);
fftw_plan plan_res = fftw_plan_dft_2d(height, width, res_freq, res_spatial, FFTW_BACKWARD, FFTW_ESTIMATE);
// Execute the FFT of the images
fftw_execute(plan_src);
fftw_execute(plan_tpl);
// Compute the cross-correlation
for(i = 0; i < fft_size ; i++)
{
res_freq[i][0] = tpl_freq[i][0] * src_freq[i][0] + tpl_freq[i][1] * src_freq[i][1];
res_freq[i][1] = tpl_freq[i][0] * src_freq[i][1] - tpl_freq[i][1] * src_freq[i][0];
tmp = sqrt(pow(res_freq[i][0], 2.0) + pow(res_freq[i][1], 2.0));
res_freq[i][0] /= tmp;
res_freq[i][1] /= tmp;
}
// Get the phase correlation array = compute inverse fft
fftw_execute(plan_res);
// Find the peak
struct Peak pk;
IplImage* peak_find = cvCreateImage(cvSize(tpl->width,tpl->height), IPL_DEPTH_64F, 1);
double *peak_find_data = (double*) peak_find->imageData;
for(i = 0 ; i < fft_size ; i++)
{
peak_find_data[i] = res_spatial[i][0]/(double) fft_size;
}
CvPoint minloc, maxloc;
double minval, maxval;
cvMinMaxLoc(peak_find, &minval, &maxval, &minloc, &maxloc, 0);
pk.pt = maxloc;
pk.maxval = maxval;
// Clear memory
fftw_destroy_plan(plan_src);
fftw_destroy_plan(plan_tpl);
fftw_destroy_plan(plan_res);
fftw_free(src_spatial);
fftw_free(tpl_spatial);
fftw_free(src_freq);
fftw_free(tpl_freq);
fftw_free(res_spatial);
fftw_free(res_freq);
cvReleaseImage(&peak_find);
//fftw_cleanup_threads(); //Cleanup everything else related to FFTW
return pk;
}
问题是与fftw_free(src_freq);
返回我一个错误(无效的指针),我只是找不到为什么...
感谢
+0
我解决了它,这是一个愚蠢的错误... 我只是忘了初始化k ... – user1977881
这里是[模板匹配(HTTP: //docs.opencv.org/doc/tutorials/imgproc/histograms/template_matching/template_matching.html)来自OpenCV文档的教程。 – flowfree