문서 영상의 적응적 이진화를 위한 국부적 water drop 기법

초록

This paper proposes an adaptive binarization using local water drop on document image. Input image is classified character and background by global threshold and background is replaced white level. Classified image is divided into blocks and according to mean of each block, water dropping or not is decided. The number of white level of each block determines volume of water adaptively. The block, which is decided water drop, is applied mask to find minimum values of each block. After water dropping, the volume of dropped water is calculated by the difference between classified image. Finally, binary image is turned out due to the volume of water Using Otsu's binarization method. Simulation results show that proposed method is more speedy and more efficient than existing methods.