Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

How to port image registration C++ code to Python

The python code below should register an image. The code is based on the example code map_test.cpp found here and also seen below. However I cannot seem to find the equivalent for casting map_ptr (which is a reg_Map object) to a map projection. The cpp code uses a .get() function. Performing this step is relevant as the normalize function is a class member of cv2.reg_MapProjec. In the code below, when I try to cast mat_ptr to map_proj, I get the error TypeError: Expected Ptr cv::UMat for argument '%s'. Does any know how to solve this issue?

Thanks in advance

def calc_homography_pixel(img1, img2):

    mapper = cv2.reg_MapperGradProj()
    mapp_pyr = cv2.reg_MapperPyramid(mapper)
    map_ptr = mapp_pyr.calculate(img1, img2)

    map_proj = cv2.reg_MapProjec(map_ptr) # the issue is how to the map from the pyramid cal.
    map_proj = map_proj.normalize()

    dest = map_proj.inverseWarp(img2)

The c++ code

static void calcHomographyPixel(const Mat& img1, const Mat& img2)
    static const char* diffpixel = "Difference pixel registered";

    Ptr<MapperGradProj> mapper = makePtr<MapperGradProj>();
    MapperPyramid mappPyr(mapper);
    Ptr<Map> mapPtr = mappPyr.calculate(img1, img2);

    MapProjec* mapProj = dynamic_cast<MapProjec*>(mapPtr.get());
    cout << "--- Pixel-based method\n" << Mat(mapProj->getProjTr()) << endl;

    Mat dest;
    mapProj->inverseWarp(img2, dest);
    showDifference(img1, dest, diffpixel);