带几何约束的彩色图像选择性分割

为了解决彩色图像分割任务中有选择性的定位感兴趣区域的具体需求,基于Lavdie-Chen的灰度图像单水平集选择性分割方法,提出带几何约束的彩色图像选择性分割方法。该试验方法将彩色图像作为一个整体,求其梯度及边缘检测函数,借助边缘检测函数、目标物体约束点确定的距离函数以及形成的多边形内外面积,共同决定曲线演化进程中的方向与速度。区域信息的加入克服了边缘函数依赖单一图像梯度的缺点;正则化优化算法的引入克服了凹陷处分割效果不理想的缺点;加法分裂算子算法可以快速求解模型的Euler-Lagrange方程。试验结果表明,本研究提出的彩色图像选择性分割方法具有有效性强和正确性高的特点。

Color image selective segmentation under geometrical constraints

In order to solve the specific requirements of selectivity during the course of color image segmentation, an active contour-based color image segmentation method under geometrical constraints was proposed based on the gray image selective segmentation using one level set by Lavdie-Chen. A color image was treated as a whole for the gradient and the edge detection function. The velocity and direction of the curve evolution were determined by the edge detection function, the distance function defined about a set of points near the boundary of the interested region and the inner and outer polygon areas of the given points. Region information could help to overcome the drawbacks of edge functions relying on a single image gradient; the regularization algorithm was introduced to overcome the shortcomings of the poor segmentation effect in the depression;the Euler-Lagrange equation was quickly solved by the additive operator splitting method. Experimental results showed that the proposed color image segmentation method had the characteristics of high validity and high accuracy to selectively segment the wanted region.