一种基于双尺度高斯核方向导数的图像轮廓检测算法

Image Contour Detection Using Double-scale Gaussian Directional Derivatives Filter

为了降低图像轮廓检测中纹理对检测结果的影响,提出一种基于双尺度高斯核方向导数滤波器的图像轮廓检测算法。结合大小两个尺度高斯核方向导数滤波器构造图像的边缘强度映射(ESM),小尺度高斯核方向导数滤波器增强了图像细节信息的检测能力,而大尺度高斯核方向导数滤波器起到抑制纹理的作用。利用ESM自身的全局和局部信息对ESM进行均衡化。通过阈值化和形态学处理,得到最终轮廓检测结果。实验表明,该方法有效可行。

In order to reduce influence of texture on the image contour detection results,an image contour detection algorithm based on double-scale gaussian directional derivatives filter is propsoed.Integrating a large scale factor gaussian directional derivatives filter with a small scale factor one to create the edge strength map（ESM）of an image,in which the utilization of small scale factor gaussian directional derivatives filter improves the detection performance of detail of image and of large scale factor one suppresses the texture.The equalization of the ESM is done using its global and local information.The final contour detection results are obtained through the operator of thresholding and morphology.Comparison with classical edge detection algorithm,Canny algorithm,the proposed method not only preserves the detection performance of image contour but also suppresses partial texture.