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.