Identify text-free PDFs for batch OCR?

I'm not sure it's possible. There was some discussion earlier in the year: