Sonar image denoising and segmentation techniques based on neutrosophic set

Due to the complex marine environment and the inherent characteristics of imaging equipment, sonar images often suffer from issues like speckle noise, weak edges, and shadows. To overcome these challenges, a noise reduction and segmentation technique for sonar images based on the neutral set domain is proposed. Firstly, the advantages of the Non-Subsampled Contourlet Transform are combined with the clustering capabilities of the neutrosophic set to process high-frequency sub-band noise. This approach effectively removes noise from sonar images while preserving edge information. Next, the neutrosophic set is utilized as a “transformation” function, and the concept of fuzziness is introduced to enhance the brightness and edge details of the true subset images. This enhancement improves the visibility of target objects in the sonar images. The self-adjusting spectral clustering technique is then applied to partition the true subsets into multiple independent subregions. By extracting only the brightness and narrowness features of these subregions, precise segmentation of target objects is achieved. The effectiveness and robustness of this proposed method are validated through the denoising and segmentation of sonar images containing various target objects. Importantly, it does not lead to missegmentation in shadowed or other interfering areas. In summary, this technique addresses the challenges of noise and segmentation in sonar images by combining NSCT, neutral sets, and spectral clustering, resulting in improved image quality and accurate target object segmentation.