基于形状共享初始化主动轮廓模型的图像分割方法

Shape sharing initialized active contour model for image segmentation

就初始轮廓提出了一种新的基于形状共享思想的初始轮廓学习方法。利用不同种类的物体全局形状或局部形状可能相似的现象,首先提取测试图像的局部形状;再找出样本库中与其局部形状相匹配的局部形状集;根据测试图片与样本图片中局部形状的相对位置及大小,进行全局形状映射;最后依照全局形状的覆盖率分组,融合成一系列初始形状。将这一系列的初始轮廓作为主动轮廓模型的初始迭代函数。另外,该主动轮廓模型结合了测试图像的边缘信息与区域信息,利用彩色梯度表示边缘的变化。从实验结果可以看出,将学习到的初始轮廓加入混杂主动轮廓中能包含更丰富的形状信息,可获得更准确的分割结果,收敛速度更快。

This paper proposed a novel shape sharing based method for initial shapes learning. The main insight of the method was that objects between different categories often share local shapes. To exploit the shape sharing phenomenon,it firstly extrac- ted the local shapes of the test images. Then it found the matched local shapes set in the exemplar database. The object shapes from the exemplar database were subsequently transferred to the test image based on the size and relative location of the local shapes. Finally, it could obtain the initial shapes in accordance with the global shape coverage. It regarded these initial shapes as the initial involution functional of the active contour model. In addition, the active contour model integrated the boundary of the color gradient with the region information. The results show that this scheme of initial shape learning can express the shape information more efficient and the segmentation results are more accurate.