彩色图像中所含有的颜色信息和纹理信息量很大且非常复杂，伴随着多个目标区域的出现，当前图像分割法很难整合纹理特征与颜色特征，图像分割效果不佳，因此，提出一种基于图割框架的改进多层图彩色图像分割方法，将多层图彩色图像分割问题看作是计算机视觉中的二元标识问题，转换成能量函数最小化问题，给出多标签MRF马尔科夫随机场能量函数，将颜色信息与纹理信息进行融合，通过建立多层图割模型对多类彩色图像能量函数的最小化问题进行求解。为了得到多类分割结果，采用一种自适应迭代收敛准则进行修改，在图割框架下进行多次迭代分割。实验结果表明，所提方法不仅分割效果良好，而且性能优越。%The color image contains massive and complex color information and texture information,which appear with the advent of multi⁃target area. It is difficult for the available image segmentation methods to integrate the texture feature and color feature,and their image segmentation effect is poor,so an improved color image segmentation method of multilayer graph based on graph cut framework is put forward. In the method,the color image segmentation problem of multilayer graph is regarded as the binary identification problem in computer vision and converted into the energy function minimization problem. The multi⁃la⁃bel MRF Markov random field energy function is given to fuse the color information and texture information. The multilayer graph cut model was established to solve the minimization problem of various color image energy functions. In order to get vari⁃ous segmentation results,an adaptive iteration convergence criteria is adopted to improve repeatedly iterative segmentation by means of the graph cut framework. The experimental results show that the proposed method has good segmentation effect,and su⁃perior performance.
展开▼
