曲率驱动扩散图像边缘形态复合滤波方法仿真

Simulation of Edge Shape Compound Filtering Method for Curvature Driven Diffusion Image

针对当前方法图像边缘滤波效果差、滤波耗时长的问题,提出基于自适应阈值的曲率驱动扩散图像边缘形态复合滤波方法。利用高斯算子微分构建曲率驱动扩散图像边缘形态检测算子,得到图像边缘角度,并对输入的曲率驱动扩散图像卷积,获得图像边缘角度方向的图像边缘响应。利用图像边缘角度的数量等条件计算图像的梯度,通过设定滤波器的尺寸来分析图像边缘滤波的抗噪性能;在此基础上,利用图像边缘的最大候选阈值计算图像的高、低阈值,并对图像的梯度做归一化处理,利用图像灰度级的中心距计算相对的峰度和偏度,并对灰度级相对的图像类间方差展开计算可以得到候选图像边缘阈值,通过对图像阈值的自适应选择,最终实现对曲率驱动扩散图像边缘的形态复合滤波。实验结果表明,提出的方法在对图像边缘形态复合滤波时,不仅具有较好的滤波效果,所用的滤波耗时也较短。

Due to poor filtering effect of image edge and long filtering time,this article puts forward a method of compound filtering for image edge shape in curvature-driven diffusion image based on adaptive threshold.Firstly,differential Gaussian operator was used to construct the detection operator of edge shape of curvature-driven diffusion image,and the angle of image edge was obtained.Then,the input curvature-driven diffusion image was convoluted to obtain the image edge response on the angel of image edge.The gradient of image was calculated by the number of image edge angles.Moreover,the anti-noise performance of image edge filtering was analyzed by setting the size of filter.On this basis,the maximum candidate threshold of image edge was used to calculate the high and low thresholds of image and the gradient of image was normalized.After that,the center distance of image gray level was used to calculate the relative kurtosis and skewness and the candidate image edge thresholds was obtained by calculating the inter-class variance corresponding to gray level.Through the adaptive selection of image threshold,the morphological compound filtering of curvature-driven diffusion image edge was achieved.Simulation results show that the proposed method not only has better filtering effect,but also has shorter filtering time during the compound filtering of image edge shape.