Research And Application Of RGBD Image Matting Using Depth Assisted Active Contours

Image matting refers to estimate foreground information of the target object accurately,thus can separate the foreground and background of the image.The traditional image matting method is combined with the use of color images and addition information,which core points can be summarized as follow: 1)the generation of the addition information;2)the refinement of the edge.However,the method only based on color information cannot solve the situation which the background is similar to the foreground.With the development of 3D reconstruction,depth map has attracted people's attention,and the distance information provided by it can distinguish the target object with difference distances.Based on this,this paper proposes a depth assisted active contour matting method.The main contribution of this paper can be summarized as following:(1)Research of depth map repairing algorithmDepth map is a distance image captured by a depth camera.The distance information can be used to distinguish the target when the background is similar to the foreground but the depth values are different.However,the depth map obtained in practice is often missing,and there are some holes.Therefore,the depth map needs to be repaired before use it.(2)Research of depth map assisted active contour and Level Set methodOne of the core points of image matting is to generate addition information.This paper use the depth-assisted active contour and level set method to segment the image.Compared with other segment methods,the level set method can be closer to the boundary of the object to be segmented in the process of segmentation,so as to better maintain the geometric topology of the object and obtain the required binary segmentation result.(3)Research of matting and the fusion of the matting resultIn the final stage,the proposed method will obtain the matting results for color information and depth information respectively,then fuse the color map results and the depth map results based on the given discriminant rules.Make the edge part of the final obtained result more in line with the actual situation.(4)Application of this methodThe source of the research is that the 3D reconstruction method has the problem that the noise in the process of obtaining the 2D image leads to the poor reconstruction model.In practical application,the obtained image opacity information is regarded as the mask and calculated with the original image to obtain the image with noise removal,then the final object model will be obtained in the further 3D reconstruction process.The experimental results show that the proposed method can better repair the depth map,which can obtain better binary segmentation results in the segment process of depth-assisted active contour and level set method.Finally,obtain the matting result that is more in line with the actual situation.