新颖的物体轮廓提取：基于双边滤波的多级边缘检测*

提出了一种新颖的物体轮廓提取方法,即通过多级边缘检测来提取物体的主要边界。双边滤波器用来建立多级,同时Canny边缘算子相应地产生边缘图,组合边缘图构造出一幅多级图。次段被定义为边缘像素的连接结构,提取并连接这些次段可以构成闭合轮廓。最终图像中最相关的闭合轮廓被判定为真实的物体轮廓。实验结果表明,该物体轮廓提取方法具有较高的可靠性并且受噪声影响较小。

Novel object contour extraction: multi scale edge detection based on bilateral filtering

This paper proposed a novel object contour extraction method, in which major object boundaries were extracted through multi scale edge detection. The bilateral filter was used to build the scales, and the Canny edge detector was accor dingly applied to produce the edge maps which would be gathered to construct a scale map. The sub segments, which were defined as the connected structure of edge pixels, were extracted and linked to form closed contours. Eventually the most relevant closed contours were considered as the true object contours in an image. The experimental results indicate that the proposed method for object contour extraction is reliable and robust to image noise.