摘要
针对现有边缘检测算法难以提取图像任意方向的边缘特征,提出基于全向小波的图像边缘检测算法.首先,定义了全向小波的概念、构造其模型并推导了全向小波的最大值与梯度模值相等的关系.理论分析表明本文算法始终沿小波变换值的最大值方向提取边缘.然后,选择二维高斯函数实例化模型,以8方向和3×3变换窗为例进行算法设计.标准图像对比试验表明本文算法能提取更多方向的边缘特征、边界清晰度也比SADD算法、Canny算子分别高出约2.17%、8.66%.
Existing edge detection algorithms are difficult to extract image edge characteristics in all directions.An edge de- tection algorithm based on Omni-directional wavelet transform （OWT） was proposed in this paper. First, the concept of Omni-direc- tion wavelet was proposed. We constructed its model and analyzed the relation between the maximal value of Omni-directional wavelet and gradient magnitude. Theoretical analysis demonstrated that OWT always extract edges along the direction of maximal gratitude. Second,2D gauss function was selected to instantiate the model. We designed OWT algorithm based on 8 directions and 3 ＊ 3 mapping window. Experimental results show that OWT can extract plenty edge features in more directions from visual sense and better than SADD,Carmy algorithm in Image-Edge-Definition （IED） by 2.17,8.66 percent,respectively.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2012年第12期2451-2455,共5页
Acta Electronica Sinica
基金
国家自然科学基金(No.61100215
No.61173036)
湖南省科技厅科技计划(No.2011GK3200)
湖南省自然科学基金湘潭联合基金(No.12JJ9021)
湖南省高校创新平台开放基金(No.2009GK3016)
湘潭大学博士启动项目(No.10QDZ30)
关键词
梯度模值
方向小波
边缘检测
图像处理
gradient magnitude
directional wavelet
edge detection
image processing
