Abstract: Image segmentation is an important processing step，but problems always exist such that the result is not satisfied and the computational complexity is too high.A novel method of binarization for gray images is presented.The Ncut（Normalized graph cut） is used as the measure for spectral clustering in the algorithm，and the weight matrixes used in evaluating the graph cuts are based on the gray-scale image，rather than the matrix commonly used based on image pixels.Thus，the proposed algorithm requires much smaller spatial costs and much lower computation complexity.Experiments on text images in natural scene show the superior performance of the proposed method on time and system resources cost compared to the typical threshold algorithms.

Key words: binarization, graph cut, spectral clustering

摘要： 图像分割作为图像识别的一个重要处理步骤，但存在效果不理想或者计算复杂度过高的问题。提出一种新的灰度图像二值化的方法。该方法将Ncut作为谱聚类的量度，在计算该值时使用基于图像灰度级的权重矩阵，而非普通基于图像像素的权重矩阵。这样，计算复杂度和空间复杂度都明显降低。通过对实际场景中文本图像的实验，数据表明此方法在时间和系统开销方面比传统基于阈值的分割方法具有更优的性能。

关键词: 二值化, 图像分割, 空间聚类