信息保真边缘检测

在YiLuo提出的边缘保真去噪算法的基础上，给出了边缘保真去噪（EPS）的计算公式，边缘检测对图像处理是很重要的并且检测的效果在很大程度上取决于图像的信噪比和所用的滤波因子，通过对人工合成图像的EPS处理，可以得出三点有实用价值的结论：（1）滤波因子的长度必须小于图像信号的1／2周期，这意味着因子长度的选择必须满足采样定理；（2）图像中含有随机噪音和阶跃型陡边界时，应用EPS能得到很好的效果，边缘的三要求－陡度，坐标位置和跳跃幅度能够完全保持；（3）对于倾斜边缘，检测效果不是很好，而对于曲线型的边缘，用于EPS处理后其效果比常规光滑滤波方法还要差。这说明EPS的应用条件是值得重视的。

HIGH S/N RATIO EDGE DETECTION

Based on the noise reduction algorithm introduced by Yi Luo (The Leading EDGE, 2002, 21(2)), the authors present some new calculate formulas and a flow chart for edge preserving smoothing (EPS) in this paper. As we know, the edge detection is very important for image processing and its efficiency strongly depends on signal to noise ratio and smooth filter operators. By using EPS method to process the synthetic image data with different S/N and various edges one can draw the follow conclusions: (1) The length of filter operator must be less than half signal period. This means that the selection of operator length should be concordant with sample theorem. (2) People can expect good results for image with sharp step edge and random noise, and the steepness, coordinate position, step amplitude of an edge can be preserved completely after EPS processing. (3) People may not expect good results for an oblique edge and may be even worse than the conventional filtering methods if the image contains curve like edges. That means people must pay attention to applicable preconditions when EPS method is employed.