Research On Active Contour Model Driven By Fuzzy C-means Algorithm

As the basis of image processing and computer vision,image segmentation is a technology that divides an image into several non-overlapping regions according to some features,and proposes the target which is desired.It has been widely used in the field of industrial inspection,agricultural machinery,imaging medicine,intelligent transportation,aerospace,surveillance security and so on.With stronger scalability,higher accuracy and more favorable results for subsequent processing,active contour model has become one of the most popular methods for image segmentation.By transforming the abstract process of image segmentation into the intuitive process of energy functional minimization,active contour model can segment two-phase images with intensity inhomogeneity effectively.However,most existing models still have some room for improvement in terms of time spent,segmentation accuracy,robustness to initial contour,anti-noise ability and stability.This paper conducts research on this issue so that one improved algorithm and two improved models are proposed.1.Improved fuzzy c-means algorithm.In the improved algorithm,a linearly weighted sum image is utilized to replace original image as the fuzzy clustering sample,which solves the problem that the algorithm is sensitive to noise.And a suitable pre-processing procedure is added which greatly accelerates the convergence speed of the algorithm.Different from existing methods,this paper adopts the strategy of integrating the cluster centers which are the results of the improved fuzzy c-means algorithm into active contour model to achieve image segmentation.2.Improved fuzzy c-means algorithm based region model.By using the cluster centers instead of the fitting functions,and replacing the level set regularization term and length term with a sign function and Gaussian filter function respectively,this model is proposed on the basis of Region-Scalable Fitting model.Compared to the fitting functions,the cluster centers have lower computational complexity and do not need to be repeatedly updated in each iteration,which greatly accelerates the segmentation speed of the model.Because the cluster centers solve the locality problem of the fitting functions,the robustness to initial contour of the model is significantly improved.In addition,the noise reduction processes added into the fitting energy term and improved algorithm enable the model to effectively deal with various types of noise.3.Improved fuzzy c-means algorithm and adaptive functions based edge model.By replacing the coefficient of the area term,edge indicator function and evolution speed function with the adaptive sign function,adaptive edge indicator function and improved evolution speed function respectively,this model is proposed on the basis of Distance Regularized Level Set Evolution model.The adaptive sign function not only solves the problem that the curve can only move in one direction,which greatly improves the robustness to initial contour of the model,but also gives the model the ability to selectively segmenting targets.The adaptive edge indicator function can not only accurately locate the target boundary,which improves the segmentation effect on complex images,but also accelerate the evolution speed of the curve.The improved evolution speed function further optimizes the stability of the model with a better function performance.Similarly,the noise reduction processes in the model and improved algorithm enhance the anti-noise ability of the model.Moreover,this model has the characteristic of edge-based plus region-based,which can effectively deal with images with large noise interference inside targets.The effectiveness of the above research contents has been fully proved by experimental results and discussions.The comparisons with some classic active contour models have proved that the proposed models have better segmentation ability and segmentation effect.