摘要
目前纹理分类算法普遍存在分类率低、耗时长的问题,改进局部二元模式(LBP)直方图算法,引入LBPV算子,结合新的全局匹配穷尽搜索方案,可弥补LBP只提取局部旋转不变纹理特征的不足,采用基于LBP特征估测主方向和特征降维的方法,则可以大大降低计算复杂度。仿真结果表明,相对于传统的LBP特征提取算法,改进的算法具有更高的纹理识别效率。
In view of the drawback of most traditional texture classification algorithm which is time consuming and has low classification rate, a new operator--LBPV by the improvement of the local binary pattern （LBP） histogram algorithm is proposed to overcome the defects of the LBP joint variance, and a new global matching based on exhaustive search is proposed to make up for the shortcomings of LBP that extracted only partial rotation invariant texture features. To reduce the computation, the paper proposes a new method on the basis of rotation invariant LBP to estimate the principal orientations and reduce the feature dimensions, Rota- tion invariant texture classification reaches a good compromise in classification accuracy as well as classification time.
出处
《电子技术应用》
北大核心
2013年第12期138-140,144,共4页
Application of Electronic Technique
基金
国家重点自然科学基金项目(60932007)
关键词
纹理分类
局部二元模式
全局匹配
主方向估计
特征降维
texture classification
LBP（Local Binary Pattern）
global matching
estimate principal orientation
reduce feature dimensions
