Image Segmentation Algorithm And Optimization Based On Active Contour Model

The active contour model has a rigorous mathematical foundation,diverse forms,flexible structure,and superior performance.It is widely used in various image segmentation problems and has achieved remarkable results in medical,military,industrial and other fields.At present,the active contour model is still in the development stage,and there are problems such as sensitivity to initial position,poor anti-noise,and inability to extract weak boundaries effectively,resulting in poor image segmentation.The theoretical framework needs further improvement,and the application scope needs to be further expanded.Drawing on this point,we propose an improved active contour model based on the level set model and the clustering method,and verify the validity of the model.The main research contents include:The fitting of the energy function of the local binary fitting model only uses the local neighborhood information of the image,which is equivalent to the local mean filtering,so the model has a certain degree of robustness to the image contaminated by Gaussian noise,due to For the salt and pepper noise,the function of the mean filtering is useless,and the error is affected by the noise.In view of this,according to the characteristics of the local fitting term in the local binary fitting model,we adopt a median local fitting term that can avoid the effect of salt and pepper noise,and the minimum value of its functional function takes the median value of the local region.It is equivalent to making a median filter to eliminate the effect of salt and pepper noise.In order to avoid the problem that the local binary fitting model is easy to converge to the local extremum and cause the segmentation failure,the fusion distance regularization level set evolution model is adopted.These two models complement each other and effectively combine the edge information and the regional information.The respective shortcomings of the foot make the algorithm improve the segmentation rate while ensuring the accuracy.The scaling function of the kernel function of the LBF model is a fixed arbitrary value,and it is not possible to adopt different scales in different regions.The gray level uniformity of different points is different,so it is inaccurate to use only one fixed-scale model to calculate statistical information of different regions.To solve this problem,we propose an adaptive Gaussian kernel function.Different scale parameters are used in different regions of the image to make the energy difference of the image more obvious,and the segmentation accuracy of the speckle noise image is improved.In order to overcome the dependence of the local Gaussian distribution fitting energy model on the size and position of the initial contour,combined with the K-means clustering method,the initial contour is automatically obtained.At the same time,a newkernel function is adopted to avoid the inappropriate characteristics of the original Gaussian kernel.Finally,the performance of the two algorithms is tested by experiments to verify the effectiveness of the proposed algorithm.The experimental results show that the proposed method can increase the robustness of the local binary fitting model to the salt-and-pepper noise and the rate of curve evolution.The local Gaussian distribution fitting energy model can acquire the initial contour autonomously and effectively segment the target.The paper has a total of 26 pictures,7 tables,and 53 references.