双三次B-样条平滑滤波算子的研究

Research on the bicubic B-spline smoothing filter operator

边缘检测一直是图像处理中的难题 .常用的边缘检测算子有 :Canny算子、Deriche算子、Sobel算子和Prewitt算子 ,它们将边缘检测算子归结为一个恰当的平滑方法及计算滤波之后的梯度或求二阶导数的零交叉指纹图 ,没有很好地解决平滑噪声和边缘的精确定位之间的矛盾 .充分利用边缘信息的多尺度性及小波变换的特性 ,通过B 样小波变换 ,给出了双三次B 样条平滑滤波算子 ,并且利用该算子进行了大量的边缘检测实验 .实验结果表明 ,文中所给的方法对边缘检测非常有效 ,具有良好的鲁棒性 .

For the processing of an image, it is extremely difficult to find the contours of objects in a scene. The classic edge detection operators, such as Canny operator, Deriche operator, Sobel operator, and Prewitt operator, tried to tackle this problem by using an appropriate smoothing filter, computing the smooth gradient or calculating a zero crossing in the second directional derivative. But the contradiction between smoothing noises and accurate location of the edge was not well solved. In this paper, on the basos of the B spline wavelet transform, the bicubic B spline smoothing filter operator is introduced with a method for detecting the edge by using this operator presented. The experimental results show that our method is efficient in edge detection, and that it is very robust to noises.