自适应双树复小波遥感图像复原

Adaptive dual-tree complex wavelet algorithm for remote sensing image restoration

由于遥感图像先验知识难以获取，提出了一种自适应的双树复小波迭代收缩复原算法。该算法根据模糊程度和噪声程度估计正则化参数，并利用经验公式计算收缩阈值。在实际应用中，算法能有效解决两步迭代算法使用固定参数的缺点，从而达到提高图像复原质量的目的。实验表明：相对于两步迭代算法，该算法复原图像的峰值信噪比提高0．64～12．23dB，收敛速度提高1．4～16倍；同时，算法在提高图像复原质量、抑制噪声干扰及减少计算时间方面优势明显。

An adaptive dual-tree complex wavelet algorithm was proposed to solve the classical image restoration problem. This method is more suited to the situation that a priori information of remote-sensing image is hard to obtain. The algorithm estimates regularization parameter from both the blurred level and the noise level, and estimates the noise using an empirical formula. In practical applications, the algorithm can effectively overcome the drawback of the two-step iterative shrinkage algorithm due to the use of a fixed parameter, and better imagery restoration quality could be obtained. Experimental results show that the image peak SNR improves 0.64-12.23 dB and the convergence speed improves 1.4-16 times. The algorithm has apparent advantages with respect of producing better restoration results, noise disturbance suppression and the reduction of computation time.