Study On Image Segmentation And3D Reconstruction Based On Graph Cut Theory

In recent years, the graph cut theory algorithm has been widely used in the field of computer vision as to minimize the energy function optimization. Such as using graph cuts theory achieving image segmentation, the three-dimensional reconstruction of the object or scene, medical image segmentation, navigation, obstacle avoidance, path planning of mobile robot, texture segmentation, image restoration and so on. It’s advantage of being able to clearly describe the problem to be solved and being able to use a soft constraint, such as spatial correlation. It is possible to avoid a difficult decisions prematurely. Image segmentation and3D reconstruction based on graph cuts theory of become a focus of research objects. With the proposed new theories and new methods of various disciplines, it also raised a number of image segmentation techniques that combine specific theories, methods and tools, such as image segmentation algorithm based on graph cuts theory becomes a hot spot. Image segmentation and3D reconstruction method based on graph cuts algorithm and watershed transformation is proposed in this paper.Firstly, the paper discusses image segmentation and stereo vision technology, the research status, the exiting problem and development trends, and introduces the traditional image segmentation method in detail.Secondly, the image is segmented into a large number of small partitions making use of the watershed transformation. Then, a simplified network graph is constructed using the partition information. And then the graph cut algorithm is employed to realize network graph segmentation and get the segmentation results. The experimental results indicate that the proposed algorithm not only solves the watershed transformation over-segmentation phenomena, but also improves the speed and accuracy of the graph cut algorithm.Finally, the paper presents a three-dimensional reconstruction method based on the mesh candidates and graph cuts on the basis of further research on graph cut theory. The method sets up3D grid nodes in the public view of stereo camera and considers grid nodes as candidate nodes of depth. Using the similarity of candidate the depth nodes with three-dimensional grid nodes to construct a three-dimensional grid map. And then using graph cuts algorithm to minimize the energy function of the3D grid map Through the minimization of the energy function solving the global optimum, and dividing of the three-dimensional grid nodes, realizing the real scene of3D reconstruction.