盈亏修正法图象边缘检测

Image Edge Detection by Modifying Profit and Loss Data

边缘检测是数字图象处理中一种重要的处理手段 ,目前普遍采用的方法是用高斯函数或者 B-样条对原始图象进行预平滑 ,然后求其一阶导数的极值点或拉普拉斯变换的零交叉作为边缘特征点 .但是在其原始图象与平滑图象的之间的残余误差中可能存在一些边缘特征信息 .为了尽可能提取残余误差中存在的边缘特征点 ,因此利用 B-样条平滑公式 ,建立了一种盈亏修正图象边缘检测新方法 ,其原理是 ,首先对原始图象数据进行盈亏修正 ,使得原始图象与平滑图象之间的残余误差进一步减少 ,然后使用修正的数据通过 B-样条平滑公式进行零交叉边缘检测 .数值实验结果表明 ,这种方法及其同类算子具有更强的边缘特征检测能力 ,并能获得较好的效果 .

Edge detection is a important process in computer vision, currently popular method firstly smooth original image using Gaussian or B Spline, then finding the extremum of first derivative or zero crossing of second derivative as edge point. But the residual between the original image and smoothing image may include some edge feature point. In order to finding the edge feature point in residual possibly, this paper put forward a novel image edge detection method by modifying profit and loss data by B Spline smoothing formula. Its principle is that firstly we modify original image intensity for decreasing the residual between smoothing image and original image and then use modified data for zero crossing edge detection by B Spline smoothing formula. Practical numerical experimental results showed that this method have a stronger edge detection ability and produced a better effect compare with kindred method.