摘要
在基于小波变换的织物图像纹理分析中,小波基的选择和分解级数的确定对纹理参数的有效性有显著影响。为此,以织物图像经小波分解后输出的高频子图像能量最小作为逼近条件,从小波库中优选出最佳小波基。对选出的最佳小波基,采用相邻两层的高频子带能量之比值小于1作为最佳分解级数的选择依据。实验证明,选用coif5小波基对织物纹理图像进行2级分解,提取分解后7个细节图像的熵作为纹理特征参数,采用ISODATA算法聚类分析时,可取得较高的正确分类率。
The selection of wavelet base and the confirmation of decomposition progression have an obvious affect on the texture analysis based on wavelet transform. In this paper, the approximation condition of selection optimal wavelet base chosen from the wavelet base bank is the energy of the high frequency subimages exported from fabric image by wavelet decomposing. The proper decomposition progression is confirmed when the energy ratio of detail subimages in two consecutive levels is less than 1. The experiment result show that high classification accuracy is achieved using ISODATA clustering algorithm with the entropy of 7 texture parameters extracted from the subimages decomposed at level 2 with coif5 wavelet base.
出处
《丝绸》
CAS
北大核心
2008年第4期37-39,共3页
Journal of Silk
基金
江苏省"六大人才高峰"项目(苏人通[2006]174号)
关键词
织物纹理
小波基
分解级数
聚类分析
Fabric texture
Wavelet base
Decomposition progression
Cluster analysis
