基于Shearlet变换的混合图像分割方法

摘要

针对图像分割中普遍存在的过分割和不完全分割问题,提出基于Shearlet变换,综合利用自组织特征映射和脉冲耦合神经网络的混合分割方法.对输入图像进行各向异性滤波,增强图像纹理信息;利用平移不变Shearlet变换对去噪图像进行变换,得到低频子带图像和高频子带图像;从低频子带系数中提取熵和偏度纹理信息,与低频子带系数组成特征向量集,利用自组织特征映射网络进行初步分割;利用改进的脉冲耦合神经网络进一步处理,减少初步分割结果的错分割现象,得到最终的分割图像.遥感图像和医学图像的分割结果表明,与传统分割方法相比,提出方法的分割准确性显著提高,具有更好的抗噪性能.%Aiming at the problem that there is universally over-segmentation and incomplete segmentation in most of image segmentation methods,a mixed image segmentation method was proposed based on Shearlet transform and combination of self-organizing feature map (SOFM) and pulse-coupled neural network (PCNN).The input image was filtered using an anisotropic filter to enhance the texture feature.It was separated into low and high sub-band frequencies using the shift invariant Shearlet transform (SIST).Feature vector was built from the low frequency coefficients complemented with texture information extracted from the low frequency coefficients,and SOFM was used for the preliminary classification using the feature vector.The modified PCNN was used to augment the SOM results to reduce the error-segmentation artifacts.For remote-sensing image and medical image,experimental results show that the proposed method is superior to the traditional method and it has better anti noise performance.