基于t混合和SMEM算法的彩色图像分割

Color Image Segmentation Based t Mixture Model and Split and Merge EM

采用t混合模型建立图像的颜色,纹理及空间位置特征的联合分布,及改进的分裂—融合EM算法(SMEM)估计混合模型的参数,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改进的SMEM算法对于数据的初始化不敏感,能收敛到全局最优,且能自适应的的选择分割的数目,因此该方法能取得更好的分割结果。

In this paper,we establish a joint color-texture-position features space using t mixture model.The parameters of t mixture are estimated by modified SMEM algorithm,image is segmented according to Bayes minimization error principle.T mixture model is robust and modified SMEM algorithm is less sensitive to initial parameters,may be converged to the global optimum and adaptive select the number of segmentation.The proposed method has better experimental results.