dip::EuclideanSkeleton doesn't work (Unhandled exception)

Question

I am trying to make a dip::EuclideanSkeleton. However, when executed, it throws me an Unhandled exception. Is it a mismatch between the return type and the type passed to the function? If so, how can I fix this? I have no more guesses. Without dip::EuclideanSkeleton the program works.

---

#include <iostream>
#include <opencv2/core/core.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <string>
#include "diplib.h"
#include "diplib/file_io.h"
#include "diplib/regions.h"
#include "diplib/display.h"
#include <dip_opencv_interface.h>
#include "diplib/binary.h"
 

int main() {
    cv::Mat img = cv::imread("Cables.jpg", cv::IMREAD_GRAYSCALE);
    if (img.empty()) {
        return -1;
    }
    cv::Mat thresh = img.clone();
    medianBlur(thresh, thresh, 7);
    threshold(thresh, thresh, 127, 255, cv::THRESH_BINARY);
    dip::Image image = dip_opencv::MatToDip(thresh);
    dip::Image conv = dip::EuclideanSkeleton(image);
    cv::Mat final = dip_opencv::DipToMat(conv);

    cv::imshow("", final);
    cv::waitKey(0);
}

in:

•   dataType_   {dt=UINT8 (1) } dip::DataType
  

out:

•   dataType_   {dt=SFLOAT (9) } dip::DataType
  

Unhandled exception at 0x00007FF85A0F3B29 in Skel_endpoints.exe: Microsoft C++ exception: dip::ParameterError at memory location 0x000000BAF730EDD8.

Answer 1

The problem is that dip::EuclideanSkeleton expects a binary image, but OpenCV knows no binary type. thresh is of type 8-bit unsigned integer, and when converting to a DIPlib object, the conversion cannot know whether this is meant to be an 8-bit uint image or a binary image. Furthermore, OpenCV uses values of 0 and 255 to represent binary images, whereas DIPlib uses values of 0 and 1. This is described in the documentation.

The solution is to force the 8-bit uint image to binary. The simplest approach is simply comparing to zero: dip_opencv::MatToDip(thresh) > 0. If you do this, you can skip the call to cv::threshold.

Next, the conversion dip_opencv::DipToMat(conv) will produce a CV_8U image, but containing values 0 and 1, not 0 and 255 as expected by other OpenCV functions. Again, one simple approach is to just multiply the result by 255: dip_opencv::DipToMat(conv) * 255.

The program would look like this:

#include <opencv2/core/core.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/highgui/highgui.hpp>
#include "diplib.h"
#include <dip_opencv_interface.h>
#include "diplib/binary.h"


int main() {
   cv::Mat img = cv::imread("erika.tif", cv::IMREAD_GRAYSCALE);
   if (img.empty()) {
      return -1;
   }
   cv::Mat thresh = img.clone();
   medianBlur(thresh, thresh, 7);
   dip::Image image = dip_opencv::MatToDip(thresh) > 127;
   dip::Image conv = dip::EuclideanSkeleton(image);
   cv::Mat final = dip_opencv::DipToMat(conv) * 255;
   cv::imwrite("foo.png", final);
}

---

It is unfortunate that MSVC doesn't automatically show the what() string for the uncaught exceptions. I recommend that you put a try/catch block around all your code so that you can display this string. DIPlib gives useful information in exceptions thrown:

int main() {
   try {
      // your code here
   } catch (std::exception const& stde) {
      std::cout << "Caught exception:\n" << stde.what() << '\n';
   }
}