Image Binarization Based on Canny's Operator

We propose a binarization algorithm based on Canny's operator after an investigation to the drawbacks of the classic binarization algorithms, Ostu algorithm and Bernsen algorithm. The algorithm uses both edge and intensity information, and generates the binarization image by filling the image, which was binarized using a high threshold, with seeds near edges of the Canny edge image, and remedying it with the image binarized using a low threshold. This method overcomes the drawbacks in the classic binarization algorithms: noise sensitivity, edge coarseness, fake shadow etc. The experimental results show that the method is effective on the binarization of low contrast images and the images in which object intensity is not uniform.