Wavelets-based Denoising by Optimizing Polynomial Threshold Function

Abstract

In this paper, a wavelets-based technique for denoising of one-dimensional signals is proposed which employs artificial bee colony (ABC) algorithm for optimizing a new polynomial threshold function. The coefficients of the threshold function are optimized dynamically to maximize the output signal to noise ratio (SNR). ABC algorithm calculates the coefficients for maximum output SNR. The proposed technique is tested for three artificial signals, namely blocks, bumps and Doppler signals. Additive white Gaussian noise (AWGN) is added to the signals and the results of denoising are presented. The results are compared with hard, soft and non-negative garrote threshold functions for three methods, VisuShrink, SUREShrink and MinMax, of determining threshold at different levels of decomposition. The performance of the algorithm is compared using different parameters such as input and output SNR, cross correlation coefficient and mean square error. The results show that the proposed algorithm provides better denoising results in all respects.

Keywords: Denoising, wavelet thresholding, polynomial threshold function, artificial bee colony algorithm