My BrightEdges edge detection algorithm is now part of OpenCV open source contributions https://lnkd.in/eESub_d It reveals edges that other algorithms often missed. An example is located here https://lnkd.in/e8Q5JW6
Here are a few examples of results
With a chess board, algorithm computes the absolute difference or two blurs with different kernel size, resulting in a edges being local minimum. Then a contrasting algorithm checks for local minimum and line continuity.
Different contrast value can be applied in a range of 1 to 255 (grayscale).
If you want the version with noise correction get it from here https://github.com/fiammante/opencv_contrib/blob/brightedges2/modules/ximgproc/src/brightedges.cpp
and the demo with webcam side by side comparison of canny and brightedges is here https://github.com/fiammante/opencv_contrib/blob/master/modules/ximgproc/samples/brightedgesexample.cpp
Machine Learning png, with the difference, then default contrast, and noise reduction applied.