16 bit image crashes Image.ThresholdBinary in EMGU.CV

Question

I have two versions of this app below. This one, and a rewrite I wrote to work with 8 bit images. 8 bit images work fine. 16 bit images cause Image.Threshold to crash in the native OpenCV call for threshold, the only error returned:

A first chance exception of type 'Emgu.CV.Util.CvException' occurred in Emgu.CV.dll Additional information: OpenCV:

The below causes a crash on my "hand coded" 16 bit image, it also crashes on a 16 bit TIF image, loaded via

var img = new Image<Gray,UInt16>("test.tif");

The crashes are identical, with identical error messages.

Here's my app:

Thoughts?

using System;
using System.Runtime.InteropServices;
using Emgu.CV;
using Emgu.CV.Structure;
using Emgu.CV.UI;
using Microsoft.VisualStudio.TestTools.UnitTesting;

namespace MyUnitTest
{
    [TestClass]
    public class UnitTestOpenCV
    {
        // Set up an image with a solid rectangle in it.
        private IntPtr _smallImg;
        private const int Ncols = 500;
        private const int Nrows = 100;

        [TestInitialize]
        public void Init()
        {
            const int n = Ncols * Nrows;
            var pixels = new short[n];
            _smallImg = Marshal.AllocHGlobal(n * 2);
            Marshal.Copy(pixels, 0, _smallImg, pixels.Length);
            for (int col = 75; col < 75 + 200; col++)
            {
                for (int row = 25; row < 25 + 50; row++)
                {
                    Marshal.WriteInt16(_smallImg, 2 * (row * Ncols + col), 10000);
                }
            }
        }

        [TestCleanup]
        public void Cleanup()
        {
            Marshal.FreeHGlobal(_smallImg);
        }

        [TestMethod]
        public void FindingMinBoundingRectangle()
        {
            var img = new Image<Gray, ushort>(Ncols, Nrows, Ncols, _smallImg);
            var rotatedImg = img.Rotate(20, new Gray(0), false);
            var bwImg      = img.ThresholdBinary(new Gray(5000), new Gray(255));
            var contour    = bwImg.FindContours();
            var rect = contour.GetMinAreaRect();
            rotatedImg.Draw(rect, new Gray(30000), 50);
            rotatedImg.Save("img.tif");
            AOIAlgorithmsBase.SaveImage("myImg.tif", img.MIplImage.imageData, Ncols, Nrows, Ncols);
            ImageViewer.Show(rotatedImg, "Image Viewer Window");
        }
    }
}

Answer 1

Actually even if EmguCV reads 16-bit images, most processing routines require 8-bit or 32-bit images. From OpenCV docs for the threshold method:

Parameters:
src – input array (single-channel, 8-bit or 32-bit floating point).

In my case, in convert my images to 8-bit for all the Emgu processing and displaying, and I convert back to 16-bit for saving in tif format.