由于传统算法无法消除相关噪声对于水下于图像的影响,通过最大后验估计构造了一个具有凸函数性质的变分模型用于抑制水下图像的相关噪声。在变分模型中通过贝叶斯约束项限定噪声的概率分布;采用马尔可夫约束项保证恢复图像满足空间连续性;通过梯度法获得变分模型的最优解,即不含相关噪声的恢复图像。实验结果表明,该算法能够有效地消除相关噪声,恢复图像的主观视觉效果较好。
The traditional algorithm cannot get rid of the influence of correlated noise on underwater images. In view of this,this paper proposes a restoration algorithm to restrain the correlated noises of underwater images by establishing a variational model based on the theory of maximum a posteriori estimation that represents convex property. In the variational model,the probability distribution of noise is guaranteed via the item of Bayesian restraint. Meanwhile the property of spatial continuity is satisfied via the Markov constraint. The gradient method is employed to obtain the optimal solution of the variational model,i. e. the restoration image without correlated noise. Experimental results show the correlated noise is eliminated by the proposed algorithm effectively; the restoration images present an excellent visual effect.