Hi, I need to implement a mechanism that is able to clean up the image of a "document" taken on a fairly uniform background - such as a top surface of a desk. Also, the document would be surrounded by background on all sides. I am trying to figure out a mechanism by which I can detect the document boundary. I am considering techniques where I can use the knowledge that the document would be surrounded by background on all sides. I'd appreciate any advice - I've tried the following
- Apply a blurring filter followed by Canny edge detection
- histogram backprojection
None of them seem to work satisfactorily.
Regards, Kashyap