The light field camera can record both direction and light intensity information of the scene from multiple viewpoints at the same time,so the light field image has larger amount information and stronger plasticity.The depth optimal estimation of light field image has gradually become a research hotspot with the development of light field technology,but multi occlusion is still a technical difficulty that has not been solved yet.Therefore,an improved depth estimation method of light field image for multi occlusion scenes is proposed.In this method,Canny edge detection algorithm with optimized high low threshold ratio is used to extract the edge of occluded area.An iterative weighted update operator with acceleration is introduced into the traditional AP clustering algorithm,which effectively improves the accuracy of clustering and solves the multi occlusion problem for depth estimation of light field.A matching cost is used to estimate the initial depth,and the initial depth results are optimized using Markov random field to obtain the final depth estimation results.The experimental results show that the proposed method not only resists multiple occlusions significantly,but also obtains lower MSE values and better depth estimation results compared with the existing depth estimation method based on light field image.