关键词: 二值化, 直方图, 优化, 分割, 阈值

Abstract: Unlike the usual way to find out automatically a threshold from image histogram, the image binarization is dealt with as an optimization problem where the best threshold could be found out by minimizing a weighted function of error sum of squares, with a fast iterative optimization algorithm given for the purpose. The new approach is compared with a most classic and commonly-used binarization method, and the result shows that the two methods are equivalent in view of either theoretical analysis or experimental data. However, our formulation relevant to the problem enables us to derive up a much more efficient algorithm available to more applications, especially to real-time video monitoring and tracking systems.