In order to enhance the robustness and segmentation performance of the fuzzy local information C-means clustering algorithm, a fuzzy local information C-means clustering algorithm based on feature selection is proposed. Based on the existing fuzzy local information C-means clustering algorithm, the local neighborhood constraints of the target function are added with neighborhood membership constraint, and the idea of feature selection is introduced to the target function. KL-divergence is taken as regularization factor to obtain a new robust fuzzy clustering algorithm. The local median filter is applied to the membership degree of each pixel in the end of the proposed algorithm iteration, and then the maximum membership criterion is used to realize the pixel classification, so as to obtain the final segmentation result. Experimental results show that the proposed algorithm has better segmentation performance and anti-noise robustness than that of the existing fuzzy local information C-means (FLICM) algorithm.