earticle

영어

In this paper, we proposed an adaptive threshold de-noising method by combining wavelet entropy and inter-scale correlation. Different from the traditional wavelet threshold based de-noising methods, our method can be divided into three steps. First, we decompose the noisy signal by using discrete wavelet transform (DWT), calculate the value of inter-scale correlation of the decomposed wavelet coefficients, and delete the high frequency coefficients which are smaller than the value of inter-scale correlation. Secondly, we equally divide the processed high frequency coefficients into several subintervals, calculate the wavelet entropy of each subinterval, and decide the threshold of high frequency coefficients by combining wavelet entropy and adaptive threshold rules. Finally, we de-noise the high frequency coefficients by using logarithmic smoothing threshold function, and reconstruct the ultrasonic signal. Experiment results have shown that the proposed method is better than some other de-noising methods in terms of SNR (signal noise ratio) and SDR (signal distortion ration).