The author describe an automatic threshold computation method called Minimum Error
Image Thresholding Algorithm and the implementation he made and compare it to the Otsu's one.

The source code is provided as well as the input images.

I've been able to download, build and run the code with the test program. I haven't been able to validate the result produced, as the author doesn't provide the output images. I would suggest checking output image in the test. This can be acheived easily with itkTestDriver with the following lines:

FIND_PROGRAM(ITK_TEST_DRIVER itkTestDriver)
SET(TEST_COMMAND ${ITK_TEST_DRIVER} --add-before-env PATH ${CMAKE_BINARY_DIR})

ADD_TEST( Test ${TEST_COMMAND} ${CurrentExe}
          ${CMAKE_SOURCE_DIR}/images/input.png out.png
          --compare out.png ${CMAKE_SOURCE_DIR}/images/test.png
)

The code is built on the same model than ITK's Otsu filter, and the specific parts are well described, so it is easy to use.

The code is quite nice and easy to read and should run without problems on any platforms supported by ITK.

Automatic threshold method are very useful in many domains. Having many of them make possible to chose the best one for a segmentation task. Also, this method is a reference in automatic threshold computation. It's a must have for method comparisons.

Just a few comments:

• Otsu filter use a strange convention used only in that filter in ITK: the background in white and the foreground black. You should prefer the one used elsewhere: the background in black and the foreground white


• ITK's naming convention requires to avoid abbreviations, so MinError should be replaced by MinimumError


• an enum would be easier to represent the distribution used, and would make easier the implementation of another distribution (if possible) than the m_UseGaussian and m_UsePoisson vars


• check the output images as suggested before

great contribution!