A novel binarization system for degraded document images is proposed in this paper. It combines two local adaptive binarization methods (Niblack method [1] and improved Palumbo method[2]) which are complementary to each other. Niblack thresholding algorithm behaves very well in character areas but causes a lot of noise in background areas while the improved Palumbo thresholding algorithm causes little background noise but sometimes it over-thresholds the characters. The proposed system which combines both Niblack thresholding and Palumbo thresholding takes advantages of these two methods. Experiments show that it performs very well with little background noise and can be used to binarize degraded document images in many application circumstances.