基于改进自生成神经网络的皮肤镜黑色素细胞瘤图像分割

SGNN-based segmentation for dermoscopy melanoma image

针对皮肤镜黑色素细胞瘤图像,提出一种基于自生成神经网络(self-generating neural net-work,SGNN)的自动分割算法。算法首先采用区域生长的方法将图像进行粗分割;然后将每一个子区域看作一个叶节点,根据节点之间的相邻关系定义连接规则;最后采用SGNN对这些节点进行聚类,完成黑色素细胞瘤图像的分割。本文方法克服了传统SGNN算法对样本训练顺序敏感的缺陷,提高了效率,实验结果表明,该方法能够自适应确定聚类数目并准确分割黑色素细胞瘤图像。

Aiming at dermoscopy melanoma image,an automatic segmentation method based on improved SGNN(self-generating neural network) is proposed in this paper.According to this algorithm,the image is coarsely segmented through region growing firstly;Then every sub-region obtained from region growing is taken as a leaf node and the connect rules are defined according to the neighborhood relative between nodes;Lastly,these leaf nodes are classified through SGNN based on the connected rules and the image segmentation i...