【摘要】 侧扫声呐图像受混响效应影响导致侧扫声呐图像斑点噪声强,边缘模糊,纹理较弱,严重时还会掩盖海底地貌。噪声方差是许多侧扫声呐图像变换域去噪算法的必要参数。指出了侧扫声呐图像在乘性噪声的影响下灰度值溢出的问题,并且以侧扫声呐图像中乘性噪声为背景,考虑灰度值范围对乘性噪声的抑制作用,提出了一种基于弱纹理块的噪声估计方法。算法主要根据噪声的散射模型,将侧扫声呐图像经过幂变换和对数变换,将服从瑞利分布的乘性斑点噪声变换为高斯白噪声,基于变换图像的梯度协方差矩阵和弱纹理块的动态选择,以迭代的方式确定噪声方差。实验结果表明:该算法能够去除灰度值溢出现象对噪声估计的影响,在高亮区域及背景区域的噪声估计结果稳定准确。更多还原

【Abstract】 Affected by reverberation effect,strong speckle noise,blurry edges and weak textures can be caused in side-scan sonar images,and if seriously the real seabed geomorphology will be covered up. Noise variance is a necessary parameter of the transform domain denoising algorithm for side-scan sonar images. By using multiplicative noise in side-scan sonar image as background and taking the inhibitory action of the gray value range on the multiplicative noise into account,an algorithm of noise level estimation based on weak textured patches is proposed. This algorithm is that the side-scan sonar images are first to undergo the power and logarithmic transformations based on the scattering model of noise,and then the multiplicative speckle noise that obeys the Rayleigh distribution is transformed into white Gaussian noise. Based on the gradient covariance matrix and the dynamic selection of weak textured patches of the transformed images,the noise variance can be determined in an iterative way. The experimental results show that by using this algorithm the influence of the gray value overflow on the noise level estimation can be removed and the noise estimation results are stable and accurate both in the highlight and the background areas.更多还原