PURPOSE: A method for extracting a Korean caption based on a connected object without respect to a size using a morphological feature and a decision making tree of Korean are provided to generate a connected object of an edge based on color information of a caption, thereby extracting a Korean caption without respect to a size. CONSTITUTION: An area including a caption is extracted(S2). An edge with high possibility of having the caption is extracted by selecting edges with edge correspondence pairs(S3). An edge connected object is generated. A connected object including the caption is selected(S4). A caption and a background used as input of an OCR are binarized in an image(S5).
展开▼
