活动轮廓的演化控制及其在铁谱图像分割中的应用

Evolution control of active contour and its application to ferrographic image segmentation

中文关键词  活动轮廓  水平集方法  图像分割  无迹卡尔曼滤波器

英文关键词  Active Contour Evolution  Level set method  Image segmentation  Unscented Kalman filter

中文摘要

      为了提高活动轮廓(active contour, AC)对边缘特征局部极小值的搜索效率, 从而提高其对铁谱图像的分割 速度, 提出了一种基于活动轮廓评价和演化行为控制的图像分割方法. 首先, 设计了一种基于矢量图的边缘指示函 数(edge indicator, EI)的计算方法, 相应的计算结果为活动轮廓模型建立了一个边缘指向更加明确的边缘指示 场(edge indicator field, EIF). 其次, 设计了曲线EI值的无迹卡尔曼滤波模型, 并基于此提出了活动轮廓边缘特征的跟 踪和评价方法. 最后, 根据以上评价结果调整曲线模型的参数以控制其演化行为. 这种参数调节机制保证了曲线模 型参数在不同的区域具有不同的参数设置. 试验结果表明, 该算法显著地提高了控制演化过程的灵活性以及活动轮 廓的收敛速度, 并且它能够实现对各种形状磨粒的准确分割, 不仅避免了弱边界区域的泄漏现象, 而且能够有效滤 除背景中的各种噪声干扰和非磨粒目标.

英文摘要

      In order to improve the efficiency of searching for the local minimum of edge features along an active contour (AC), and to speed up the segmentation of ferrographic images, a novel method is proposed based on the edge feature assessment and the evolution control of the AC. Firstly, the edge indicator (EI) function is calculated according to vector-valued images, and based on the calculation, the corresponding edge indicator field (EIF) is constructed for the active contour model (ACM) with more distinctive edge features. Secondly, an unscented Kalman filter of the AC’s EI is designed, and the filter is used to track and assess the AC’s edge features. Finally, based on the tracking and assessment process, the AC’s model parameters are adjusted to control the evolution of the AC adaptively. The parameters adjustment ensures that the ACM has different parameter setting in different image regions. Experimental results demonstrate that the proposed method significantly increase the convergence rate and control flexibility of active contour evolution model. And the application to ferrographic images shows the algorithm can accurately segment wear particles of different shapes. It not only avoids the leakage in weak image edges but also effectively removals false foreground, interference and noise.