基于小波阈值去噪的收缩函数改进方法

摘要

Classic threshold filtering is composed of the soft threshold and hard'threshold. For the soft shrink-function, the filtered signal is excessively smooth. For the hard shrink-function, the filtered signal associated with the Gibbs phenomenon in the signal mutation after filtering. The contraction function presented is between soft and hard. Through its wavelet coefficients modification, the filtered signal reach a reasonable balance between excessively smoothness and edge oscillation. Based on the analysis, the PSNR comparison values and the image using different filtering methods are provided. The comparative results show that the improved method elimi-nate the oscillation phenomena and retain some details effectively while remove noise from image.%经典的小波去噪方法有软阈值、硬阈值滤波两类.软阈值收缩函数滤波后信号过于平滑,而硬阈值收缩函数滤波后在信号突变处伴有吉布斯现象.因此,提出介于软、硬阈值之间的一种收缩函数,通过其对小波系数的估计,使滤波后的信号在过分平滑与边缘振荡现象之间达到合理的平衡.在分析的基础上,给出了各种方法滤波后的PSNR对比值以及图像.实验结果表明,该方法在去噪的同时既有效地消除了振荡现象,又保留了部分细节信息.