基于循环平移和DTCWT的声呐图像滤波方法

声呐图像散斑噪声的存在严重影响声呐图像的人工判读和自动识别。声呐图像散斑噪声滤波是声呐图像处理领域的一个重要难题。双树复小波变换(DTCWT)因其具有近似的平移不变性、良好的方向选择性已被应用于图像滤波。但这种平移不变性的近似性使得DTCWT不能较为彻底地抑制伪吉布斯现象,而导致滤波后图像的边缘存在一定程度的模糊。为了减小伪吉布斯现象和声呐图像边缘模糊,将循环平移与DTCWT结合使用,提出一种基于循环平移和DTCWT的声呐图像滤波方法。实验结果表明,该方法在滤除声呐图像散斑噪声的同时能够较好地保持声呐图像的边缘。

Sonar image filtering method based on cycle shift and DTCWT

The speckle noise in a sonar image seriously disturbs the manual interpretation and automatic identification of the sonar image. Speckle noise filtering of a sonar image is a critical difficulty in the field of sonar image processing. DTCWT (dual tree complex wavelet transform) has been widely applied in image filtering due to its approximate shift invariance and excellent directional selectivity. However, the approximation of this kind of shift invariance makes the DTCWT not completely inhibit the pseudo Gibbs phenomenon, which leads to a certain degree of blurriness on the edges of the filtered sonar image. In order to reduce the pseudo Gibbs phenomenon and the edge blurriness of the sonar image, this paper proposes a filtering method for a sonar image based on cycle shift and DTCWT. The experiment results show that the proposed method is able to filter the speckle noise and better preserve the edges in a sonar image simultaneously.