Sonar Image Filtering And Segmentation Based On Varational Method

Accurate and fast sonar image segmentation method meets the needs of marine development and military development.The purpose of sonar image segmentation is to classify pixels and divide sonar image into valuable areas with certain rules on the premise of minimizing the loss of image information.Because of the peculiarity of imaging,the image quality is generally not high.The method of sonar image segmentation has been studied and developed for more than ten years.Among the many methods of sonar image segmentation,variational method is widely used due to its easy modeling process,good expansibility and simple implementation.The variational method still has great research and application value in the application field and development of sonar image segmentation.Filtering is the premise of sonar image segmentation,and segmentation is a key step in sonar image processing.The results of filtering have a great impact on the results of sonar image segmentation.Because the accuracy of segmentation has a great impact on the discrimination of target objects,it is of great research value for the filtering and segmentation of sonar image.The main research content of this paper is the application of variational method in sonar image filtering and segmentation:(1)Classical filtering methods include Lee filtering,Wiener filtering,median filtering,etc.,but these methods blur the edge details while removing the noise.Based on PCA(Principal component analysis,PCA)combined with an improved L1 norm regularization adaptive total variational noise suppression model,the combination of more features can be extracted to the sonar image information,and have mature algorithm can optimize the minimizing functional,therefore this article applies the L1 norm of based on improved adaptive total variation model compared with the traditional variational model can keep the edge information,greatly after removing noise compared with the classical sonar image filtering method has a better peak signal-to-noise ratio(PSNR)and structural similarity(SSIM).The correctness of the method is verified.(2)After the sonar image filtering can undertake object detection and segmentation,the classic segmentation model includes Canny operator,Otsu operator,Log operator,variational model,the GVF model the Snake model,Mumford Shah model,C-V model,etc.,so this article on the basis of C-V model is put forward based on variational method improved Hyperbolic active contour model(from the active contour model,H-C-V model),the second,the level set function method is applied to the model,In this way,the model can be continuous in the segmentation process and also adapt to the topological changes.Finally,in order to improve the speed of curve evolution,the paper adopts narrow band method to make the active curve evolve rapidly.Compared with the classical sonar image segmentation model,this model can retain the edge information to a large extent.Experiments show that the proposed method is robust and fast in image segmentation.Edge extraction based on improved directional variational model is applied to small size targets.By analyzing the experimental results,it can be seen that the model can effectively segment the target,and the segmentation result is relatively ideal.