US Patent 7636467 Binarization of an image

For binarizing an image, which is composed of pixels, the image is split into two semi-images. Local parameters are initialized based on values of pixels that lie in an area adjacent to a boundary separating the semi-images. A binarization is then performed separately for each of the semi-images using an adaptive threshold, wherein the adaptive threshold is calculated for each of the semi-images proceeding from the initialized local parameters.