摘要
由于经典的小波阈值函数存在一定的缺陷,如硬阈值函数在阈值处不具有连续性,软阈值函数的小波估计系数和原系数之间存在着恒定的偏差,会导致去噪后的图像出现失真、产生吉布斯震荡等问题。文中综合典型的小波阈值函数的优点,并综合一些改进的方法,针对其缺点,提出了一种改进的阈值函数。该函数不仅在阈值处是连续的、小波估计的系数渐进原系数,并且具有可微性,易于实现梯度算法的自适应学习。为了验证该阈值函数的优越性,通过仿真实验对几种小波去噪方法的均方差(MSE)和峰值信噪比(PSNR)进行了对比。实验结果表明,用改进后的阈值函数进行去噪,无论是在视觉效果上,还是在均方差和峰值信噪比性能分析上均优于常用的阈值函数。
As the classical wavelet thresholding function has certain defects,for example,the hard threshold function is not continuous at the threshold,and there is constant deviation between the original coefficient for soft- threshold function,which can cause image distortion after denoising and produce the problem such as Gibbs phenomena. An improved threshold function based on the advantages of the typical wavelet thresholding function and combined some improved methods is proposed. The function is not only continuous at the threshold,the estimated wavelet coefficients approaching the original coefficient,but also differential and easy to realize the adaptive learning of gradient algorithm. In order to verify the superiority of the thresholding function,through the simulation experiment,the Mean Square Error（ MSE） and Peak Signal- To- Noise Ratio（ PSNR） from several wavelet denoising methods are compared. According to the experimental results,this proposed method has better in visual effect and performance analysis for MSE and PSNR than the traditional threshold functions.
出处
《计算机技术与发展》
2016年第5期76-78,共3页
Computer Technology and Development
基金
江苏省自然科学基金(BK2011789)
东南大学毫米波国家重点实验室开放课题(K201318)
