Abstract. To solve the non-uniformity of micro-CT image with CV(Chan-Vese) model and the influence of location of initial contour curves on segmentation speed in the LGIF(Local and Global Intensity Fitting) model, K-LGIF(K-means-Local and Global Intensity Fitting) model was proposed through adding K-means clustering information into energy function of LGIF active contour model. The K-LGIF model extracts outline of the image as the initial contour to reduce the number of iterations and shorten time consuming. Comparing measured geometry parameters by simulating symptoms of osteoporosis and normal mouse femur of trabecular bone and using gray level co-occurrence matrix, we measured the parameters of texture distribution of trabecular bone. The experimental results show that the K-LGIF model can effectively improve segmentation of non-uniform gray image and increase speed of segmentation. This method may provide an approach for the quantitative analysis of osteoporosis.