Abstract: A multi-scale edge detection in the EMD domain using the median absolute deviation to estimate the noise threshold was proposed to detect the edges of noisy signals with different signal to noise ratios. The new algorithm first calculates the slope of the EMD residuals in scales. Second, it eliminates the noise in the slope signal by thresholding, the noise threshold of which is estimated by the median absolute deviation. At last, it outputs the edge information by the spatial consistence test Simulation results show that, the multi-scale edge detection in the EMD domain using the adaptive noise threshold can accurately detect the edges of signals, and suppress the noise.

Key words: Edge detection, Empirical mode composition method, Multi-scale, Median absolute deviation