Occlusion Handling System Based On Depth Image

Augmented reality technology develops based on the virtual reality technology,which adds virtual objects into the real world,expanding and enhancing the real world.Augmented reality technology has been widely used in medical,mechanical,military and other fields.Occlusion handling in augmented reality affects the fusion effects of real world and virtual objects.Therefore,solving the problem of occlusion handling is of great significance to the development of augmented reality applications.Based on the analysis of the existing methods,we construct an occlusion handling system with better effect combined with the matting algorithm.Firstly,the color image and the depth image are simultaneously acquired by the Kinect.Then we preprocess images.Secondly,through depth comparison and expansion,the original depth image is processed into trimap which is one of the inputs to the matting algorithm.We select the appropriate matting algorithm such as shared matting,and then improve it with depth information.The improved shared matting algorithm has higher accuracy when dealing with scenes with similar foreground and background colors.Through the matting algorithm we obtain the foreground with high quality edges and transparency channels.Finally,compose the foreground and virtual objects.The results of occlusion handling system are evaluated by qualitative and quantitative methods.The proposed method is compared with the direct depth comparison,joint bilateral filter and guiding filter.The results show that the proposed method has a higher accuracy.The processing speed of the proposed occlusion system depends on the size of the occlusion area.In most augmented reality applications,the system can process in real time because the occlusion area is small.