Visual Saliency와 해리스 코너 검출기를 이용한 자막 영역 추출 기법

Abstract

Texts inserted in TV contents help viewers understand information visually. A text extraction approach is used for video information retrieval and indexing. Most of text detection and extraction methods are based on color and contrast information between texts and backgrounds. In this paper, we propose a text region extraction method using DCT(Discrete Cosine Transform) and Harris coner detector. The proposed method consiss of three steps: (1)detecting candidate text regions by DCT technique, (2)detecting candidate text regions by Harris corner detector, (3)extracting final text regions by combining two kinds of the candidate text regions. Experimental results showed that the proposed method extracts the text regions effectively.