【摘要】 多小波变换系数在同一尺度上具有多个输出块,传统的多小波系数相关去噪算法只利用了尺度间或同一尺度上相同块内相邻变换系数之间的相关性;为了改善去噪效果,根据白噪声和信号的多小波变换系数的不同特性,将同一尺度上不同块的变换系数做相关运算及归一化后分别与原来块的变换系数进行比较,根据相应的规则对变换系数进行取舍,然后再采用"相邻系数法"对保留的变换系数进行处理,进一步地去除噪声。由于同时考虑了同一尺度上不同块之间和相同块内相邻变换系数之间的相关性,仿真和实验结果表明该方法能够更好地减少噪声对小波变换系数的干扰,突显出信号突变点的小波变换系数,具有更好的去噪效果。更多还原

【Abstract】 When a signal is decomposed by multiwavelet,the multiwavelet coefficient consists of several blocks,and the correlation of coefficients in same blocks or cross the scales is utilized to denoise the signal usually.According to the different characteristics of signal and white noise,we put forward a denoising algorithm based on the correlation of multiwavelet coefficients for the better denoising performance.Firstly,we calculated the correlative coefficient of different blocks.Then,we compared the correlative coefficient with normalized correlative coefficient and made selection according to the rule,in which the reserved correlative coefficient was processed with neighboring coefficient method.Using the algorithm,the correlation of neighboring coefficients and different blocks was taken into account simultaneously.So the denoising performance was better than that of the conventional algorithm which only adopted the neighboring coefficients.The experimental results show that the interference of noise is reduced,and the break point is more visible.更多还原