Aiming at shortcomings of the geodesic active contour（ GAC） and Shi level set segmentation methods

an improved strategy for liver CT images is proposed. Image preprocessing technology is used to realize automation of initial curve of algorithm. Use weighted median and average gradient to improve the GAC algorithm

and introduce the Poisson function to improve the Shi algorithm. The ground truth image of the liver region after segmentation is obtained by using the connected domain extraction technique

the image filling technique and partial manual method. Results of the experiment

based on peak signal-to-noise ratio

interactive information and segmentation error rate to evaluate the segmentation effect

show that the improved level set method not only effectively enhances the accuracy of liver region segmentation

but also reduces the time complexity of the algorithm.