您可以删除无用的行和列,并使用原始矩阵一半大小的矩阵工作。
resize
您可以使用最接近的插值函数轻松完成此操作:
#include <opencv2/opencv.hpp>
#include <iostream>
using namespace cv;
using namespace std;
int main(int argc, char **argv)
{
Mat1b mat = (Mat1b(4,4) << 0, 1, 2, 3,
4, 5, 6, 7,
8, 9, 10, 11,
12, 13, 14, 15);
Mat1b res;
resize(mat, res, Size(0, 0), 0.5, 0.5, INTER_NEAREST);
cout << "Mat:" << endl << mat << endl << endl;
cout << "Res:" << endl << res << endl;
return 0;
}
然后中的值res
只是您需要的索引处的值:
Mat:
[0, 1, 2, 3;
4, 5, 6, 7;
8, 9, 10, 11;
12, 13, 14, 15]
Res:
[0, 2;
8, 10]
为了将值恢复到原始位置,您可以使用具有合适模式的 Kronecker 产品(在 OpenCV 中不可用,但可以轻松实现)。这将产生:
Mat:
[0, 1, 2, 3;
4, 5, 6, 7;
8, 9, 10, 11;
12, 13, 14, 15]
Res:
[0, 2;
8, 10]
Res Modified:
[1, 3;
9, 11]
Restored:
[1, 0, 3, 0;
0, 0, 0, 0;
9, 0, 11, 0;
0, 0, 0, 0]
代码:
#include <opencv2/opencv.hpp>
#include <algorithm>
#include <iostream>
using namespace cv;
using namespace std;
Mat kron(const Mat A, const Mat B)
{
CV_Assert(A.channels() == 1 && B.channels() == 1);
Mat1d Ad, Bd;
A.convertTo(Ad, CV_64F);
B.convertTo(Bd, CV_64F);
Mat1d Kd(Ad.rows * Bd.rows, Ad.cols * Bd.cols, 0.0);
for (int ra = 0; ra < Ad.rows; ++ra)
{
for (int ca = 0; ca < Ad.cols; ++ca)
{
Kd(Range(ra*Bd.rows, (ra + 1)*Bd.rows), Range(ca*Bd.cols, (ca + 1)*Bd.cols)) = Bd.mul(Ad(ra, ca));
}
}
Mat K;
Kd.convertTo(K, A.type());
return K;
}
int main(int argc, char **argv)
{
Mat1b mat = (Mat1b(4, 4) << 0, 1, 2, 3,
4, 5, 6, 7,
8, 9, 10, 11,
12, 13, 14, 15);
Mat1b res;
resize(mat, res, Size(0, 0), 0.5, 0.5, INTER_NEAREST);
cout << "Mat:" << endl << mat << endl << endl;
cout << "Res:" << endl << res << endl << endl;
// Work on Res
res += 1;
cout << "Res Modified:" << endl << res << endl << endl;
// Define the pattern
Mat1b pattern = (Mat1b(2,2) << 1, 0,
0, 0);
// Apply Kronecker product
Mat1b restored = kron(res, pattern);
cout << "Restored:" << endl << restored << endl << endl;
return 0;
}