摘要
针对小波变换的局限性和人类视觉系统的敏感性不同,提出了一种新的基于小波包与人类视觉系统的边缘检测算法.首先对加噪的载体图像应用中值滤波进行图像的平滑处理并用小波包对图像的高频和低频部分进行分解;然后根据人类视觉系统模型的特性进行小波包系数的选择,并利用Log算子进行边缘的零交叉点检测;最后通过二维小波包重构函数得到图像近似部分的边缘.理论数据与实验结果表明,应用文中提出的边缘检测算法,通过与经典的Roberts、Sobel、Prewitt和Log 4种边缘检测算子比较,不但更符合人类的认知,而且在含有丰富细节和微小变化的区域,新方法更具有应用价值.
To recognize the limit of wavelet transform and the sensitive diversity of human visual system, an approach using a new edge detection algorithm based on wavelet packet and human visual system is proposed. Firstly, the median filtering is easy to remove image noise, and the denoising image is transformed by wavelet packet;The edge detection regions are by selecting the wavelet packet coefficients based on the characteristic of human visual model ,and then use Log algorithm to select the zero-crossing;The detected edge image is obtained by reverse wavelet packet. The experimental results show that the algorithm compared with Roberts ,Sobel ,Prewitt and Laplacian of Gaussian edge detection algorithms not only agrees more with human' s recognition, but also can get an even better application on some regions including abundant local details and some tiny areas that change.
出处
《南京师大学报(自然科学版)》
CAS
CSCD
北大核心
2010年第4期134-138,共5页
Journal of Nanjing Normal University(Natural Science Edition)
基金
山东省自然科学基金(ZR2009GM009)
山东省博士基金
关键词
小波包
人类视觉系统
中值滤波
小波包系数
零交叉点
wavelet packet, HVS, the median filtering, the wavelet packet coefficients, zero-crossing