改进的GrabCut算法木材表面缺陷图像分割试验

Image Segmentation of Wood Surface Defects by Improved GrabCut Algorithm

试验采用传统GrabCut算法和改进的GrabCut算法,针对单目标、多目标、复杂背景下多目标的木材表面缺陷图像进行多组对比实验。结果表明:改进后的GrabCut算法,针对木材表面的缺陷图像分割进行了优化,能有效改进传统GrabCut算法中的欠分割和过分割、易受区域凹凸纹理的干扰等缺点,而且分割各类木材表面缺陷图像时都能取得较好的效果。说明改进后的GrabCut算法具有其优势和可行性。

The traditional GrabCut algorithm and the improved GrabCut algorithm are used to simulate the multi-target wood sur- face defect images in single target, multi-target and complex background. The improved GrabCut algorithm was optimized for the defect image segmentation of the wood surface, which could effectively improve the defects of the traditional Grab- Cut algorithm, such as under-segmentation, over-segmentation, and susceptible to interference from regional bumps. The segmentation of all kinds of wood surface defects images achieves better results, and the improved GrabCut algorithm is with its advantages and feasibility.