一种改进的图像分割活动轮廓模型

AN IMPROVED IMAGE SEGMENTATION ACTIVE CONTOUR MODEL

C-V模型具有计算复杂度低、对初始化和噪声不敏感等优点,在处理图像的时候总是从全局的角度去考虑图像区域的灰度变化,从而导致难以分割灰度不均的图像。局部二元拟合(LBF)模型在处理灰度不均匀的图像分割方面有很大优势,但是LBF模型存在依赖初始轮廓大小、位置等缺点。针对C-V模型不能分割灰度不均图像和LBF模型敏感于轮廓初始化的问题,给出一个用偏微分方程表示的新的融合局部(LBF模型)和全局信息(改进的C-V模型)的活动轮廓模型。实践结果表明,新的模型对初始轮廓的敏感性低,能分割灰度不均的图像,且优于C-V模型,其分割效率明显高于LBF模型。

Chan-Vese （C-V） model has the advantages of low computational complexity and insensitive to initialisation and noise, however when processing the image, it always deals with grey change of image region from global perspective, thus leads to difficult in segmenting the images with intensity inhomogeneity. Local binary fitting （LBF） model has great advantage in dealing with the segmentation of images with uneven greyscale, but it has the shortcomings of relying on the initial size and position of contours. In light of the problems that the C-V model cannot segment the image with intensity inhomogeneity and the LBF model is sensitive to contour initialisation, we propose a new active contour model expressed with a partial differential equation, which integrates both the local （LBF model） and global （improved C-V model） information. Experimental results show that the new model is less sensitive to initial contour so is able to segment the image with intensity inhomogeneity, and is superior to C-V model, it is also clearly better than LBF in segmentation efficiency.