Research of Optimized Denoising and Precise Segmentation Methods in Sonar Image Processing

Sonar image processing plays a crucial role in various fields such as military operations, oceanographic surveys, medical imaging, etc., where the quality of sonar images directly impacts the accuracy of subsequent data analysis and decision-making. To identify the most suitable denoising and segmentation methods for sonar image, this paper delves into the image denoising and image segmentation issues of sonar images for the image quality improvement through a comprehensive comparison and evaluation of various classical algorithms. In the denoising domain, we conducted a detailed analysis of commonly used algorithms including mean filtering, median filtering, Gaussian filtering, morphological filtering, and Fourier filtering, and objectively evaluated their performance using metrics such as peak signal-to-noise ratio and mean squared error. Experimental results indicate that median filtering excels in noise reduction in sonar images, demonstrating significant improvement in image clarity and quality. For segmentation, we compared and studied K-means clustering, maximum entropy method, OSTU's method, and OSTU's method based on connected component analysis. Through quantitative evaluation of segmentation effectiveness and qualitative analysis, we found that the OSTU algorithm based on connected component analysis performs exceptionally well in segmenting sonar images, accurately extracting target regions and laying a solid foundation for subsequent target recognition and tracking. This work introduces a comprehensive evaluation method, providing a reference for algorithm selection in the field of sonar image processing with practical and generalizable value.