基于同化核分割相同值区域特征融合的边缘检测

针对SUSAN(Smallest univalue segment assimilating nucleus)边缘检测算子在强噪声环境中容易产生虚假边缘的不足,提出了基于同化核分割相同值(Univalue segment assimi-lating nucleus,USAN)区域特征融合的边缘检测方法——FUSAN(Fusion-USAN)算子。该算子在利用USAN区域大小信息提取潜在边缘点的基础上,分别直接和间接利用USAN区域的形状信息计算边缘响应,并采用简单的融合策略融合这两种边缘响应得到最终的边缘响应。与SUSAN算子相比,FUSAN算子综合利用了USAN区域的大小和形状信息。利用合成图像和常用Cameraman图像并加入噪声对FUSAN算子进行评估,实验结果表明:FUSAN算子的检测性能优于SUSAN算子。

Edge detection based on feature fusion of USAN area

The Smallest Univalue Segment Assimilating Nucleus(SUSAN) edge detector is prone to false-edge problem in noisy environment since it only utilizes the size of USAN area.To solve this problem,a novel edge detector based on feature fusion of the USAN area is proposed,which is called FUSAN edge detector.First,the FUSAN edge detector finds candidate edge points using the size information of USAN area.Then,directly and indirectly using the shape information of USAN area,the FUSAN edge detector computes two edge responses.Finally,the two responses are fused using simple strategies to get the final edge response.Comparing to SUSAN edge detector,FUSAN edge detector utilizes both the size and shape information of USAN area.Synthetic and Cameraman images are used to evaluate the performance of FUSAN edge detector.Experimental results show that FUSAN edge detector outperforms SUSAN edge detector.