Based upon the maximum fuzzy partition entropy principle, a novel approach for image segmentation was presented. After the concept of fuzzy partition was introduced briefly, a new definition of fuzzy partition entropy was proposed. A threshold selection approach from gray level histogram through maximizing the entropy of fuzzy partition was put forward. It was demonstrated that KSW entropic thresholding method is just a special case of the approach proposed herein. The experiment was conducted on three real object images. The results show that the proposed approach has better performances than some classical threshold selection methods do.