基于Bandelet变换的图像去噪算法研究  

Image Denoising Based on Bandelet Transform

作　　者：蔡恬 

机构地区：[1]汕头职业技术学院计算机系,广东汕头515078

出　　处：《南方职业教育学刊》2014年第1期92-96,共5页Journal of Southern Vocational Education

摘　　要：Bandelet变换是一种信号稀疏表示方法。与二维小波变换无法有效利用图像几何特征不同的是,Bandelet变换通过计算图像的几何流并在每个局部区域内选择最佳方向实施小波变换。本文分析了Bandelet变换的基本原理并将Bandelet变换应用于图像去噪,提出一种基于Bandelet变换的图像去噪算法并在MATLAB中实现,结果验证了本文算法比小波变换具有更好的性能。本文算法的去噪图像更加平滑,视觉效果更佳,峰值信噪比(PSNR)、等效视数(ENL)等评价指标也更好。Bandelet transform is a kind of sparse representation for signal. Different from 2D wavelet transform that is not able to utilize the geometric features of an image, Bandelet transform calculates the geometric flows of an image and then applies wavelet transform in the optimum direction within every local area. The basic theory of Bandelet transform is analyzed in this paper firstly. Follow that, Bandelet transform is employed for image denoising. An algorithm based on Bandelet transform for image denoising is proposed and implemented with MATLAB. The experimental result shows that this algorithm is superior to wavelet transform for image denoising. The denoised image of this algorithm is smoother. It also has better visual effect and indicators such as Peak Signal to Noise Ratio（PSNR） and Equivalent Number of Looks （ENL）.

关 键 词：信号稀疏表示 BANDELET变换 几何流 图像去噪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]