The new Canon cameras use a hybrid focusing system that creates post-processing difficulties with Magic Lantern's RAW video output. The focus pixels end up in the image. You need to interpolate around them. There is Java based software "Pink Dot Remover" that does this. However, it has issues, the first being in Java ;)

There are multi-OS raw2dng converters that works well in creating DNG files from ML RAW.

I would like to do the following with libRaw, in VS C#, read a DNG file, interpolate (average of left and right at that chroma channel) around pixels (I know co-ordinates), then save back to a DNG file.

I've tried dcRaw "bad pixel" routines, but their interpolation is not perfect for these pixels.

I am happy to share my code/experience in this endeavor for an article. There seems to be no easy way to read/write DNG or cDNG files.

Thanks for your time!