中文摘要：

声呐图像背景复杂，对比度差，边缘恶化，不易判读图像边缘．对声呐图像执行小波变换能够有效去除噪声，但是由于小波的局限性，其对图像边缘的保持效果不佳．正交有限Ridgelet （FRIT）变换能够有效克服小波变换在处理高维信号时的不足，是一种有效处理二维奇异性信号新方法．针对水下声呐图像特点及噪声分布情况，提出单尺度自适应去噪处理算法，既削弱了单独使用FRIT处理时出现的环绕现象，又克服了单独使用小波去噪时无法保持边缘的缺点，同时提高了图像信噪比．与其他经典方法及单独使用FRIT方法比较，此改进方法在水下声呐图像去噪输出信噪比及边缘保持等方面均获得理想效果．

英文摘要：

Sonar images usually have complex backgrounds, low contrast, and indistinct edges, making it difficult to determine the edges of objects. The wavelet transform, though it can effectively eliminate noise from the sonar image, does not preserve edges in sonar images very well because of its own limitations. The finite Ridgelet transform （FRIT） is a new high dimensional signal processing method that overcomes the defects of wavelet transform in dealing with high-dimensional signals. To deal with the special charac- teristics of sonar images and the distribution of underwater noise, this paper proposed a new adaptive monoscale FRIT denoising method, which not only reduces the effects of surrounding phenomenon, but also overcomes the shortcoming of wavelet denoising, that it cannot keep edges distinct. Therefore the signal-to-noise ratio is substantially improved. Compared with other methods, this method is effective at denoising while preserving edges in sonar images.