I have a friend who is launching a new project. He wants to use some kind of OCR to detect and translate kanji characters into other languages. He hit a little brick wall in search of available algorithms to do this, because these characters are a bit more complicated than the English characters we are used to.
We suggested that he start learning about 2D convolution and Fourier transforms to begin the pattern recognition process, but he is looking for a good starting point.
Unfortunately, my knowledge of OCR is extremely limited, so any suggestions I can get through are likely to be most helpful!
source
share