To solve the problem of text segmentation in complex scene images, a method of complex scene text segmentation is proposed. The original image is firstly divided into some small homogeneous regions by using the simple linear iterative clustering(SLIC) algorithm, and the graph model is constructed based on the region neighborhood connection diagram. Then, Gaussian mixture models(GMMs) and support vector machine(SVM) post probability based model are proposed to make model for foreground(text), and the degree of each region’s fitness to models is introduced to calculate likelihood energy. In addition, to improve the discrimination ability of GMMs, a model performance descriptor is introduced to estimate parameters of GMMs adaptively. Experimental results show that the proposed method can deal with the problem of complex scene text segmentation efficiently, and the recognition precision rate is improved significantly.