Research On Sonar Image Enhancement And Segmentation Methods Based On Stationary Wavelet Transform

The recognition of underwater target sonar image provides powerful support for marine exploration,and segmentation of sonar image is an important part in the process of target recognition.Due to the poor contrast of sonar image,the target region is not easy to be separated from the complex background region.Therefore,before segmentation for sonar image,the sonar image should be enhanced to highlight the target region and suppress the background region.The stationary wavelet transform has the characteristics of multi-resolution analysis and translation invariance,which has obvious advantages in image processing.In this thesis,sonar images are enhanced and segmented based on stationary wavelet transform.In the part of enhancement of sonar image,the stationary wavelet transform is combined with the nonlinear enhancement method.Firstly,the stationary wavelet transform is applied to sonar image to obtain low-frequency subband image and high-frequency subband images.Then,the improved gamma function is used to nonlinearly enhance the low-frequency subband image and according to the characteristics of each high-frequency subband images,the inverse tangent function is used to nonlinearly enhance them,respectively.Finally,the enhanced sonar image is obtained by inverse stationary wavelet transform.The results of experiment show that the comprehensive performance of the method in this thesis is better than that of the classical gray-scale linear transformation method,UM method,HE method and CLAHE method,which can suppress the noise effectively while enhancing the contrast of sonar image.In the part of segmentation of sonar image,this thesis combines stationary wavelet transform and Fuzzy C-means(FCM)clustering method to segment the enhanced sonar image.Firstly,the stationary wavelet transform is performed on the enhanced sonar image.Then,the initial clustering centers are obtained according to the grayscale characteristics of the low-frequency subband image in the stationary wavelet domain.Finally,neighborhood informations are added to redefine the objective function of FCM clustering method,and the initial clustering centers are used to segment the enhanced sonar image with the improved FCM clustering method.The results of experiment show that the comprehensive performance of the method in this thesis is better than the traditional Otsu double threshold method,maximum entropy double threshold method and FCM clustering method,which can segment sonar image efficiently and improve the segmentation accuracy.