面向多纹理特征的脑瘤图像分割方法

Segmentation method of brain tumor image for multi-texture features

为了研究医学脑瘤图像纹理特征的选取和平滑图像分割轮廓线的问题,依据纳什均衡理论,给出了纳什均衡的多纹理特征计算方法和表示公式;并在纳什均衡多纹理计算的基础上,给出了纳什均衡计算的相似区域的判断与相似区域的合并方法,提出了用于图像分割轮廓线平滑的面向多纹理特征的改进的C-V模型。脑瘤图像分割实验结果表明:本文方法与较典型的纹理图像分割方法相比,脑瘤图像分割准确率平均提高5%,验证了本文方法的有效性。

Texture image segmentation is a key technology in image processing.To study the selection of textural features and smoothing of contour lines in medical brain tumor images,Nash equilibrium multi-texture feature calculation method and expression formula are presented on the basis of Nash equilibrium theory.Then,based on Nash equilibrium multi-texture calculation,the method for judging the similarity and merging similarity regions of the Nash equilibrium calculation is given.An improved cross-validation model for multi-texture features in image segmentation contour smoothing is proposed.The image segmentation experiment shows that compared with the typical texture image segmentation method,the proposed method is more effective because its accuracy is 5%higher on average.