Text Image Detection

Question

I got halftone images made by a cheap camera and I need to make an OCR program. The main problem is noise or objects that are not text but are present in the binary image. Now I am thinking of extracting text from an image.

I need a good algorithm for this. Can you offer a really good one? For example, if the image contains black text and something like a black line, then this algorithm will select only text without a line.

Answer 1

You describe two types of noise that you want to remove. (By the way, the wikipedia page for noise reduction is not bad, see the "in images" section).

One type is odd-point noise. This is often called “speckle” or “salt and pepper” noise and is usually removed by some kind of averaging filter. There's a nice page outlining some algorithms for this in mathworks .

The second type is strings. This is more complicated, and I would not call it noise, it would depend on the type of input image. This document seems appropriate, but it is not available for free online, so you may have to buy it or go to your local university library.

, , (), , , , .