基于增量式进化策略的图像滤波算法

Image Denoising Based on Incremental Evolution Strategy

进化滤波器采用笛卡尔遗传规划算法不断更新滤波器结构,可以用于复杂噪声图像降噪。针对这种基于大样本大数据量的进化机制,论文提出增量式进化策略,在进化过程中分阶段逐步增加样本数据量。对高斯噪声、椒盐噪声以及混合噪声的滤波实验表明,增量式进化滤波器能有效降低计算复杂度,同时在抑制图像噪声、细节和边缘保护方面,优于传统的进化滤波算法。

The evolved filter employs cartesian genetic programming to update the structure of the filter and remove complicated noise. In this paper, an incremental evolution strategy is proposed for this large training samples evolution mechanism. The amount of sample data is gradually increased by stages. Experiments on gaussian noise, salt-pepper noise and the mixed noise show that the incremental evolved filter reduces the algorithm complexity. Better properties of noise attenuation, details and edge preservation are also gained in this incremental evolved filter.