Abstract: The quality of binarization algorithm of text image directly affects the character recognition performance of text extraction based on gray-level image. Compared with several common binarization algorithms, this paper proposes an improved algorithm combined histogram analysis with OTSU. By means of the interval analysis and classification in histogram information， the algorithm reduces the search range of threshold value， and utilizes OTSU to get the optimal threshold value. The experimental results show that it offers key techniques to reduce the light influence of OTSU, and is of general segmentation applicability.


Key words: natural scenes, binarization, histogram, OTSU