복잡한 배경의 칼라 영상에서 허프 변환을 이용한 글자 영역 추출 = Location of Text Region and Skew Estimation of Textlines By Hough Transform in Complex Color Images

다국어 초록 (Multilingual Abstract)

An efficient method to extract text regions in complex color images is proposed. An algorithm locates the spatial position and estimates the skew of the text lines which are present in image using the local color information and hough transform.
The color image is divided into sub-blocks. Sub-blocks are classified into a background block and a character candidate block by intensity variance and contrast. The background intensity is erased from 3×3 neighbor blocks. Similar blocks in hue and saturation are grouped into a region. Regions are refined by several heuristic parameters such as color, size and position information.
Binary image is obtained after region refinement and is segmented into connected components. Character candidate components are detected by filtering noncharacter-like components based on several heuristics. The skew of the text lines are estimated by applying the hough transform.
The proposed method have been used to locate text regions in book cover images. Experimental results are reported.