Abstract: Since digital images often display strong characteristics of hierarchical layouts, a new hierarchical tree model targeted space constraints of pixels is defined to conduct image segmentation.The whole process is accomplished by inquiring levels of pixels, making probabilistic statistics towards the hierarchical tree, and calculating the classification results of nodal points after pruning assignable rules under the condition of confidence.Space constraints are added to accelerate the generation of the Tree.Pruning of the tree represents the segmentation process and thus setup of pruning condition exerts great influence on the final result.As experiments show, hierarchical trees based on space constraints are generated more rapidly with better image segmentation results.