摘要
利用小波分析对图像进行多层分解,然后,用分解的低频系数重构图像作为模型链接权参数W的估计,再用一种最佳阈值方法估计阈值θ,最后用最大相关准则确定网络计算的迭代次数N,成功实现了图像的自动分割.实验仿真表明,该方法在模型参数自动估计的基础上避免了PCNN对图像的过平滑作用,分割图像保留了良好的轮廓和更多的细节.
In this paper,the wavelet analysis was applied to multi-layer decomposition of image.Then,it was linked to the right as a model parameter estimates of W with decomposition of the low-frequency coefficients of the reconstructed image.It was estimated an optimal threshold value of the threshold θ.The final maximum correlation criterion was used to determine network iteration times N.A successful automatic image segmentation was obtained.Experimental results showed that the method automatically estimated model parameters based on PCNN image to avoid over-smoothing effect,segmentation images retained a good profile and more details.
出处
《云南大学学报（自然科学版）》
CAS
CSCD
北大核心
2010年第6期652-656,共5页
Journal of Yunnan University(Natural Sciences Edition)
基金
国家自然科学基金资助项目(61065008)
云南省自然科学基金资助项目(2007F174M)
云南大学青年基金资助项目(2007Q024C)
云南大学研究生科研课题资助项目(ynuy200928)
关键词
脉冲耦合神经网络
参数估计
小波分析
最大相关准则
pulse coupled neural network
parameter estimation
wavelet analysis
maximum correlation criterion
