Most recent image edge detection methods are based on exploiting spatial high-frequency to localize the boundaries and image discontinuities. These approaches are strictly sensitive to noise and also their performance decrease with the increase in noise level. Due to the successful use of the Canny operator in GPR data, the Canny operator is recognized as the most desirable edge detection algorithm. The purpose of this work is to find edges of subsurface objects such as PVC pipes with the Canny operator using the non-destructive evaluation technique of GPR and utilizing À trous wavelet transform to be well suited for non-stationary data. In synthetic data, this algorithm detects the events and anomalies with high accuracy, especially for PVC pipes that are spherical in shape. On the other side, operating on real data proved the power of this method in detecting buried objects.