摘要
利用非分离的二维离散小波可把输入信号直接进行分解且小波中的4个滤波器矩阵间存在相互转换关系,提出了改进的二维离散小波算法。该算法用低频滤波器中的数据表示3个高频滤波器中的数据,使得在卷积过程中原算法重复计算的卷积被省去。采用改进算法后,计算每个小波系数时的乘法次数由原算法的l×(l-1)减少为ll×/2。再通过阈值法设置阈值进行数据压缩,使得压缩后的能量保留在99%以上,从而保证了重构信号的失真率很小且自适应地消除加在扰动信号上的噪声。仿真结果表明该方法在保证大的压缩率的情况下提高了运算速度,并对信号中的噪声干扰有很好的消除能力。
The input signals can be decomposed directly by the non-separable 2-DWT（Discrete Wavelet Transform）and the mutual conversion is realizable among the four filter matrixes in the wavelet,based on which,an advanced non-separable 2-DWT is presented. It passes over the repeated convolution calculations in the traditional algorithm by using the data in low frequency filter instead of those in the three high frequency filters. Thus the multiplication times in wavelet coefficient calculation are reduced from l×（l-1）to l×l/2. By threshold processing,the remainder energy after data compression is over 99 % ,which makes the distortion ratio of reconstructed signals smaller and the noise of the disturbance signals eliminated adaptively. Simulation results show that the method improves operation speeds with higher compression ratio and better noise elimination capability.
出处
《电力自动化设备》
EI
CSCD
北大核心
2007年第5期53-57,共5页
Electric Power Automation Equipment
