This paper puts forward a novel image edge detection method based on uniform B-spline with shape parameter by modifying profit and loss data. The original image intensity is modified for decreasing the residual error between smooth image and original image. A smooth surface of the digital image is presented by the new modified data. The edge point is detected by either computing the local extreme of the directional derivative or computing zero crossing of the second order directional derivative of the smooth surface. Experiments demonstrate that this algorithm is simple, accurate and reliable. It can wipe off the bogus edge commendably and process the digital image in real time.