Character Extraction from the Hangeul Document Image Represented by a Quadtree

쿼드트리로 표현된 한글 문서 영상에서의 문자 추출

This paper proposes a method of extracting connected strokes and characters in a Hangeul document image represented by a quadtree. Connected strokes are extracted by examining adjacency for each node of a black region to the side of east and south. Then, each character is extracted by the sizes and the relations of connected strokes. This is done by iterative vertical and horizontal manipulation. The result shows all characters are extracted in the same sequence as that of context with space and time efficiency.