OPENCV 4. 矩阵维度、通道和数据访问

获取某一坐标的值

cvGetReal1D, cvGetReal2D, cvGetReal3D, cvGetRealND; 对单通道进行访问
CVAPI(double) cvGetReal3D( const CvArr* arr, int idx0, int idx1, int idx2 );

cvGet1D, cvGet2D, cvGet3D, cvGetND; 对多通道进行访问
CVAPI(CvScalar) cvGet3D( const CvArr* arr, int idx0, int idx1, int idx2 );

多通道：某一坐标的值都由多个数值组成，如颜色图像RGB；如下：

int main()  
{ 
    float data[18] = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18};

    CvMat mat;
    cvInitMatHeader(&mat, 3, 3, CV_32FC2, data);

    for(int y = 0; y < mat.rows; y++)
    {
        for(int x = 0; x < mat.cols; x++)
        {
            CvScalar value = cvGet2D(&mat, y, x);
            printf("(%5.2f, %5.2f)   ", value.val[0], value.val[1]);
        }
        printf("\n");
    }
    getchar();
    return 0;
}

可以看出，数据在内存中的排放顺序是依次存放的；

多维数据
CvMatND，维度可以为多维

int main()  
{ 
    float data[18] = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18};
    int size[] = {2,3,3};
    CvMatND mat;
    cvInitMatNDHeader(&mat, 3, size, CV_32FC1, data);
    for(int z = 0; z < mat.dim[0].size; z++)
    {
        for(int y = 0; y < mat.dim[1].size; y++)
        {
            for(int x = 0; x < mat.dim[2].size; x++)
            {
                int pos[] = {z, y, x};              
                double value = cvGetRealND(&mat, pos); //double value = cvGetReal3D(&mat, z, y, x);


                printf("%5.2f   ", value);
            }
            printf("\n");
        }
        printf("=========\n");
    }
    getchar();
    return 0;
}

可以用cvGet1D可以访问数据，但是这种方式太慢，因此可以观察数据在内存中的位置而直接访问内存即可；

int main()  
{ 
    float data[18] = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18};

    CvMat mat;
    cvInitMatHeader(&mat, 3, 3, CV_32FC2, data);

    for(int y = 0; y < mat.rows; y++)
    {
        for(int x = 0; x < mat.cols; x++)
        {
            float * pData = (float*)(mat.data.ptr + mat.step * x + y * sizeof(CV_32FC2));
            printf("(%5.2f, %5.2f)   ", pData[0], pData[1]);
        }
        printf("\n");
    }
    getchar();
    return 0;
}

对于多维数组，有类似的结论：

int main()  
{ 
    float data[18] = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18};
    int size[] = {2,3,3};
    CvMatND mat;
    cvInitMatNDHeader(&mat, 3, size, CV_32FC1, data);
    for(int z = 0; z < mat.dim[0].size; z++)
    {
        for(int y = 0; y < mat.dim[1].size; y++)
        {
            for(int x = 0; x < mat.dim[2].size; x++)
            {
                int pos[] = {z, y, x};              
                float * pValue = (float *)(mat.data.ptr + z * mat.dim[0].step + y * mat.dim[1].step + x * sizeof(CV_32FC1));

                printf("%5.2f   ", pValue[0]);
            }
            printf("\n");
        }
        printf("=========\n");
    }
    getchar();
    return 0;
}