Active Appearance Model (AAM) is an efficient method for extracting human facial features.It includes active appearance models and the fitting algorithm.Within all kinds of fitting algorithms

the inverse compositional algorithm is one of the most efficient algorithms.However the efficiency of the fitting algorithm will drop when some other objects occlude any parts of human face.An improved AAM fitting algorithm is presented.It not only keeps the superiority of the original inverse composition algorithm but also enhances the ability of anti-jamming.The experiments show that our algorithm can enhance the robustness of the AAM fitting algorithm and keep the useful information when fitting with occlusion.