An Effective Noise Reduction Algorithm in Signal Edge Detection

Abstract

Noise as an unwanted factor always degrades the edge detection performance. Exploiting real edges under noise contaminated condition has been a challenge in the edge detection issue, especially when the noise power is high. This paper presents a robust edge detection method in a noisy condition based on the 1D wavelet transform domain as a solution for the noise problem. First of all, a new group of wavelet basis functions for the edge detection is introduced. Then, a basic ramp function is modeled by a Gaussian approximation, and edge detection according to introduced bases derived from relevant formulas is discussed. We develop the multiscale production method and present a new algorithm to reinforce real peaks and suppress fake edges. Finally, the simulation results of the edge detecting for the noisy uniform step edge are provided to evaluate proposed algorithm efficiency. The results showed that our scheme is more effective than the production of adjacent coefficients method in the low signal to noise ratio condition.