水下图像恢复的难点在于缺少海水的点扩展函数的足够信息, 而导致病态的问题。为了提高水下激光成像系统的成像质量, 提出了用粒子群优化正则化参量的盲图像复原算法。该方法结合Tikhonov正则化和改进的全变分正则化的技术特点, 使用一种交替迭代方法, 分别估计点扩展函数和估计复原图像, 同时用粒子群算法优化正则化参量。结果表明, 该方法对水下图像复原具有较好的鲁棒性, 算法收敛稳定。

Difficulties of underwater image restoration lies in lack of enough information about the point spread function of sea water which induces the ill-posed problem consequently. In order to improve the imaging quality of underwater laser imaging system, a blind image restoration algorithm based on particle swarm optimization regularization parameter was proposed. This method integrated the technique characteristics of Tikhonov regularization and the improved total variation(TV) regularization. An alternating iterative method was adopted to estimate point spread function and restored image respectively. Meanwhile, the regularization parameter was optimized by using particle swarm algorithm. After dealing with the simulation images and the actual underwater images, the results of underwater image restoration show that this method has good robustness for underwater image restoration and the algorithm is convergent and stable.