摘要
针对传统小波阈值去噪算法容易导致重构信号出现附加的振荡和阈值处理前后存在恒定偏差的情况,提出了一种新的阈值去噪算法:在第n层小波变换阈值λn两侧分别取一正值a,b,使得小波系数的绝对值在a,b处连续且一阶可导,旨在增强阈值的灵活性。实验表明,去噪效果较传统方法更加高效。同时,在此基础上,选择对比度图像增强技术来进一步提升含噪图像分割质量。通过仿真实验,运用传统方法和本文改进算法分别对附加了高斯噪声和随机噪声的图像进行处理,结果表明,本文改进算法在信噪比和峰值信噪比数据上都具有明显的优势。
A new threshold denoising algorithm is proposed to resolve the problems of traditional wavelet threshold de-noising algorithms, which may lead to additional oscillations of reconstructed signals and constant bias before and after threshold processing. In this method a and b are respectively taken on both sides of the threshold value of wavelet transform layer n so that the absolute values of the wavelet coefficients are continuous and first order derivable at a and b. Its purpose is to enhance the flexibility of the threshold. Experiments show that the denoising effect is more efficient than the traditional method. At the same time, contrast image enhancement technique is selected to improve the quality of noisy image segmentation. The simulation results show that the proposed algorithm has obvious advantages in both SNR and SNR data, by using the pass-through method and the improved algorithm, respectively, to deal with the image with Gaussian noise and random noise.
出处
《微型机与应用》
2017年第14期39-42,共4页
Microcomputer & Its Applications
基金
四川省教育厅重点科技计划项目(14ZA0330)
四川省达州市2014年科技计划项目(2014-8220)
关键词
小波变换
阈值去噪
灰度变换
信噪比
峰值信噪比
wavelet transform
threshold de-noising
gray-scale transformation
SNR
PSNR
