中文摘要：

针对声呐图像分辨率低、边缘模糊的问题，提出一种基于非下采样Contourlet网络的声呐图像超分辨率重建方法。对高分辨率声呐图像进行降质样本构建，通过非下采样轮廓波网络来学习降质样本图像的带通方向子带及相应高分辨率声呐图像的带通方向子带间的非线性映射关系，再利用映射系数重建降质声呐图像的带通方向子带，并对降质声呐图像的低通子带进行立方插值，得到高分辨率的低通子带，最后进行非下采样轮廓波逆变换。实验结果表明，该方法重建的声呐图像具有更好的边缘、细节保持效果。

英文摘要：

Aiming at the problem of low resolution and blurred edges in sonar image,a sonar image super-resolution recon- struction algorithm based on nonsubsampled contourlet network is proposed. Firstly ,the high-resolution sonar image is degraded to build the degraded sample image, and the nonlinear mapping relationships between band-pass directional subband of degraded sample image and the corresponding band-pass directional sub-bands of the high-resolution sonar image is leamt with nonsubsampled contourlet network. Then,the band-pass directional sub-band of the degraded sonar image is reconstructed using these mapping coefficients. The low-pass sub-band of the degraded sonar image is interpolated with cubic interpolation to obtain the high resolution low sub-band. Finally, the reconstructed high resolution sonar image is achieved with the inverse nonsubsampled contourlet transform. Experimental results show that with the proposed method, the reconstructed sonar image has good effects in keeping details and edges.