Datatypes
The OpenCV matrices CvMat and cv::Mat have their own type system:
| OpenCV | Datatype |
|---|---|
| CV_8S | char |
| CV_8U | unsigned char |
| CV_16S | short |
| CV_16U | unsigned short |
| CV_32S | int |
| CV_32F | float |
| CV_64F | double |
It's kind of awkward to remember all the types and even worse, there's no type-checking at compile time. So the recent OpenCV2 introduced the templated matrix cv::Mat_<_Tp>. Now imagine we are writing a highly simple function, where we want to print a matrix.
With a cv::Mat_ we can use the advantadges of templates:
#include <cv.h>
#include <iostream>
using namespace cv;
using namespace std;
template <typename _Tp>
void printMat(const cv::Mat_<_Tp>& mat) {
typedef typename DataType<_Tp>::work_type _wTp;
for(int i = 0; i < mat.rows; i++)
for(int j=0; j < mat.cols; j++)
cout << (_wTp) mat(i,j) << " ";
cout << endl;
}
void printMat(const cv::Mat& mat) {
for(int i = 0; i < mat.rows; i++)
for(int j=0; j < mat.cols; j++)
cout << (int) mat.at<unsigned char>(i,j) << " ";
cout << endl;
}
int main(int argc, const char *argv[]) {
unsigned char a[] = {0,1,2,3,4,5,6,7,8};
Mat_<unsigned char> d0 = cv::Mat_<unsigned char>(9,1,a).clone();
Mat d1 = cv::Mat(9,1,CV_8UC1,a).clone();
cout << "d0: ";
printMat(d0);
cout << "d1: ";
printMat(d1);
}
If you run the the program you'll get:
d0: 0 1 2 3 4 5 6 7 8
d1: 0 1 2 3 4 5 6 7 8
But what happens if I now choose integers instead?
int main(int argc, const char *argv[]) {
int a[] = {255,256,257,258,259,260,261,262,263};
Mat_<int> d0 = cv::Mat_<int>(9,1,a).clone();
Mat d1 = cv::Mat(9,1,CV_32SC1,a).clone();
cout << "d0: ";
printMat(d0);
cout << "d1: ";
printMat(d1);
}
If I now run the function, you'll see where it fails... The element access to the cv::Mat with unsigned char will truncate the result:
d0: 255 256 257 258 259 260 261 262 263
d1: 255 0 1 2 3 4 5 6 7
So the first function stays unchanged, because of using templates, while printMat(const cv::Mat& mat) has to be changed to:
void printMat(const cv::Mat& mat) {
for(int i = 0; i < mat.rows; i++)
for(int j=0; j < mat.cols; j++)
cout << mat.at<int>(i,j) << " ";
cout << endl;
}
For old code I write my methods in such a way:
template <typename _Tp>
void histogram(const cv::Mat& input, cv::Mat& hist, int N) {
hist = cv::Mat::zeros(1, N, CV_32SC1);
for(int i = 0; i < input.rows; i++) {
for(int j = 0; j < input.cols; j++) {
int bin = input.at<_Tp>(i,j);
hist.at<int>(0,bin) += 1;
}
}
}
void histogram(const cv::Mat& input, cv::Mat& hist, int N) {
if(input.type() != CV_8SC1 && input.type() && CV_8UC1 && input.type() != CV_16SC1
&& input.type() != CV_16UC1 && input.type() != CV_32SC1)
{
CV_Error(CV_StsUnsupportedFormat, "Only Integer data is supported.");
lf::logging::error("wrong type for histogram.");
}
switch(input.type()) {
case CV_8SC1: histogram<char>(input, hist, N); break;
case CV_8UC1: histogram<unsigned char>(input, hist, N); break;
case CV_16SC1: histogram<short>(input, hist, N); break;
case CV_16UC1: histogram<unsigned short>(input, hist, N); break;
case CV_32SC1: histogram<int>(input, hist, N); break;
}
}
General Matrix Multiplication
In Octave:
octave:1> A = [1 2;3 4]
A =
1 2
3 4
octave:2> B = [5 6;7 8]
B =
5 6
7 8
octave:3> C = [9 10;11 12]
C =
9 10
11 12
octave:4> A*B
ans =
19 22
43 50
octave:5> A'*B
ans =
26 30
38 44
octave:6> A'*B'
ans =
23 31
34 46
octave:7> alpha=2
alpha = 2
octave:8> A*B+alpha*C
ans =
37 42
65 74
In OpenCV2 this translates to:
#include <cv.h>
#include <iostream>
int main(int argc, const char *argv[]) {
float a[] = {1.0, 2.0, 3.0, 4.0};
float b[] = {5.0, 6.0, 7.0, 8.0};
float c[] = {9.0, 10.0, 11.0, 12.0};
cv::Mat A = cv::Mat(2,2,CV_32FC1,a).clone();
cv::Mat B = cv::Mat(2,2,CV_32FC1, b).clone();
cv::Mat C = cv::Mat(2,2,CV_32FC1, c).clone();;
cv::Mat D; // result
// D = A*B
cv::gemm(A, B, 1.0, cv::Mat(), 0.0, D);
std::cout << D << std::endl;
// D = A^T*B
cv::gemm(A, B, 1.0, cv::Mat(), 0.0, D, CV_GEMM_A_T);
std::cout << D << std::endl;
// D = A^T * B^T
cv::gemm(A, B, 1.0, cv::Mat(), 0.0, D, CV_GEMM_A_T + CV_GEMM_B_T);
std::cout << D << std::endl;
// D = A*B+alpha*C
float alpha = 2.0;
cv::gemm(A, B, 1.0, C, alpha, D);
std::cout << D << std::endl;
return 0;
}