In this paper, an improved threshold segmentation algorithm based on isoperimetric ratio for infrared imaging object is proposed. The segmentation weight matrix is constructed by computing the similarity among the pixies with 4 adjoining points and stored in a sparse matrix. The isoperimetric ratio is obtained after the indicator vectors are formed with 255 gray levels. The proposed algorithm selects the minimum isoperimetric ratio confined in conditions as the best partition criterion instead of the traditional minimum isoperimetric ratio. By analyzing the variation of isoperimetric ratio with the gray levels, the proposed method can find the optimum threshold to segment infrared imaging object. Experimental results show that compared with the traditional methods, the proposed algorithm can reach a higher segmentation rate and is more robust in different kinds of infrared images.