针对太赫兹全息再现像易出现边界模糊的问题,提出了一种基于进化算法优化区域生长的分割方法。首先对原始图像进行双边滤波,同时进行形态学腐蚀操作,得到区域生长的种子。其次利用遗传算法和差分进化算法进行阈值寻优,以限制区域生长。得到太赫兹全息图像的分割结果后,以平均结构相似度(MSSIM)为客观评价来衡量算法的有效性,分割结果显示进化算法优化的区域生长算法效果较好,MSSIM可达0.8以上。最后,为比较两种进化算法的寻优性能,将算法应用于可见光图像,根据图像的分割结果,得到差分进化算法在速度和寻优能力上均优于遗传算法的结论。

Terahertz holographic reconstructed images are prone to boundary blur. Therefore, this study proposes a segmentation method based on optimized region growth by evolutionary algorithms. First, the proposed method is used to perform bilateral filtering and morphological erosion on the original images to obtain the seeds of the region growth. Second, genetic algorithm and differential evolution algorithm are used to perform threshold optimization to limit the region growth. Subsequently, the segmentation results of the terahertz holographic images are obtained. Average structure similarity (MSSIM) is used as an objective evaluation for assessing the algorithm''s effectiveness. Segmentation results show that the region-growing algorithm optimized by the evolutionary algorithm has a good segmentation effect. Moreover, the MSSIM can reach 0.8 or higher. Finally, to compare the optimization performance of two evolutionary algorithms, the algorithms are applied to visible light images. According to the segmentation results of the images, it is concluded that the differential evolution algorithm is superior to the genetic algorithm in terms of speed and searchability.