为了快速更好地进行彩色图像分割，以图像的颜色、纹理及空间位置作为综合特征，基于t混合模型，提出了一种自适应的图像分割方法，即先采用贪婪的EM（Greedy EM）算法估计混合模型的参数，然后根据贝叶斯最小错误率准则对图像进行分割。由于t混合模型的稳健性和Greedy EM算法对于数据的初始化不敏感，且能收敛到全局最优，因此与其他的方法相比，不仅速度提高，而且能取得更好的分割结果。

To obtain more efficient performance, taking color, texture and position feature in the images as their comprehensive features, an adaptive image segmentation method is proposed based on t mixture model and Greedy EM. The parameters of t mixture are estimated by Greedy EM algorithm . Image is segmented according to Bayes minimization error principle. T mixture model is robust and Greedy EM algorithm is less sensitive to initial parameters therefore it may be converged to the global optimum. Compared to other methods on image segmentation, the proposed method can accompush improved efficienty and better experimental results.