提出一种基于特征矢量集的核Logistic回归方法,解决核Logistic回归的解的稀疏性问题,降低后验概率估计的计算复杂度,该方法与Markov随机场方法相结合,应用到图像分割中,在Bayes公式中,对样本条件概率的估计转换为对核Logistic回归方法的后验概率的估计,从而提出一种新的Markov随机场模型的实现方法,在对纹理图像的分割实验中得到良好效果。 A kernel logistic regression method based on eigenvector set is proposed to solve the sparseness problem of kernel Logistic regression and reduce the computational complexity of a posteriori probability estimation. This method is combined with Markov random field method and applied to image segmentation, In the Bayes formula, the estimation of the conditional probability of the sample is transformed into the estimation of the posterior probability of the nuclear Logistic regression method, and a new method of implementing the Markov random field model is proposed, which achieves good results in the segmentation of texture images .