Based On Active Contour Model For Image Segmentation

Image segmentation aims at partition the original image into sevel homogeneous characteristic regions and isolating the region of interesting. Image segmentation is the premise of image understanding and image recognition. The active-contour-model-based image segmentation methods, which combine low-level image information with high-level priori knowledge, have shown unique advantage and comprehensive applicability contrast with the traditional image segmentation methods. But the active contour model has some disadvantages such as the high complexity of the algorithm and its poor convergence to the weak edges of images. This dissertation aims to research on image segmentation with active contour models. The main works of this dissertation are as follows:Firstly, for the noise sensitivity problem and weak edges leakage problem exist in the VEF Snake model, an improved Snake model is proposed. The vector field kernel is redefined, which reduces the sensitivity between the capture range of the external forces and the change of parameter. That a dynamic force, the combination of the CEF force field with the potential force field, is constructed in order to alleviate the weak edges leakage problem. Experiments demonstrate the improved Snake model is robust to noise and can converge to the object boundary accurately.Secondly, a gradient-based hybrid LBF geometric active contour model is presented. Due to parametric ACM can not handle topological deform, the GGVF fields are generalized to the geometric ACM which has topology-adaptable ability. The model minimizes the energy by two phases. In the first phase, the LBF method is used to optimize the model based on the image region information, which could get the image global optimum information. In the second phase, GGVF force field pulls the model contours to the object boundary in the vicinity of regions when the evolvement speed of the model is slow. Since the long-range capture, GGVF force field can pull the contours lying in or out the object boundary, the model can avoid the weak edges leakage problem which is an annoyance in traditional Level Set methods. Experimentatl results show that the model is effective.Thirdly, aim at the geometric ACM segmentation based on gradient information has the drawbacks of edge leaking. This model is improved with two facets. At the first, replace the edge function with the minimal area term, which characteraristic is independent of the image edge information even if the edge in blur or disperse state, still can get the satisfied result. At the same time, the minimal area term has the global optimize characteristic, which can detect all the contours inside the object with one initial closed curve. Secondly, on the basis of the original external force field, we introduced a diffuse region gradient flow magnitude. The diffuse region force comes from image region segmentation, which can give the model a global view of the boundary information and expand the capture scope of the curve and can enhance the robustness to the noise. Experiments show that the new model can overcome the disadvantages of edge leaking.Finally, two present variational Level Set models of image segmentation are analyzed, at the same time, introduced the Laplace zero crossing edge detection model. Based on them a new variational Level Set method is proposed integrated the image information from both boundary gradient and region. This method succeeds the advantages of these two above mentioned models while overcomes their disadvantages. Meanwhile, the method adopts the AOS numerical algorithm, which can not only grarantee the stability but also improve the convergence speed of the proposed model. Experiments illustrate that the method proposed has the ability to segment images with different types and results in stable and accurate effects.The three sections of the dissertation present the image segmentation methods which aim at solving the edges leakage problem. The first section proposed the improve method which solved the edges leakage problem when segmenting the noisy weak edge images. The second section solved edges leakage problem when segmenting the multi-level gray images with low-contrast. The third section solved the edges leakage problem when segmenting the strong noisy or weak edge images.