论文部分内容阅读
本文提出了一种用二项分布序列作为滤波模板,并相继进行拉普拉斯运算和零交点检测提取图象边缘的新的边缘检测算法。推导了二项分布——拉普拉斯边缘检测算子及实现卷积运算的快速算法。即使当尺度空间常数n较小时本文提出的算子亦具有较理想的带通特性及高频抑制性能。利用均方误差最小原理对被检测点邻域九点进行平面拟合可以成功地检测零交点,适当选择拟合平面斜率的检测阀值可大大消除噪声并有效地提取出图象边缘。
In this paper, we propose a new edge detection algorithm using binomial distribution as filter template and Laplacian and zero-crossing detection to extract image edges. The binomial distribution - Laplacian edge detection operator and fast algorithm of convolution operation are derived. Even when the scale space constant n is small, the proposed operator has better band-pass characteristics and high-frequency rejection performance. Using the mean square error minimization principle, the plane fitting of nine points in the neighborhood of the tested point can successfully detect the zero-crossing point. Appropriate selection of the detection slope of the fitted plane slope can greatly eliminate the noise and effectively extract the edge of the image.