A study on the knowledge-based thinning algorithm that preserve the shape of the Korean character image

The thinning algorithm is a preprocessing technique that is an important part of image processing techniques. As the scope of its applied and related fields, such as character recognition systems, has been wider, so has its research been active. A Korean character consists of an initial phoneme, a medial vowel and a final phoneme. It has the aspect that, according to its placement, its size and shape can be changed. Because of this, and more specifically since all three elements should be combined or connected in some way to produce a character, there occurs a problem of the distorted shape in the curved and the connected areas. This problem is mainly due to the irregular thickness of an object pixel, and seriously affects the Korean character recognition system, which uses the curved and the connected shape information for its recognition variable. This article is on a knowledge-based thinning algorithm which deals with printed Korean character images. It is intended to reduce the distortion rate of a Korean character image to its minimum level. Furthermore, I believe that this algorithm not only helps to enhance the effectiveness of the character recognition system, but it can also be applied as a medical image preprocessing technique, e.g. in the computerization of the medical administration, which is being vigorously studied.<>