We presented an improved geodesic active contour model for image segmentation. This method com- bines the geodesic active contour model and the tensor voting framework. We added in°uence of tensor voting in the edge-detector function. The proposed model is more sensitive to edges. It can detect weak boundaries of de- sired objects in some images with low contrast. Using this method, we don't have to spend time to ¯nd some suit- able parameter which is necessary in the geodesic active contour model. Numerical results show the e®ectiveness.