本文为美国加州圣何塞州立大学（作者：Manish Kumar Singh）的硕士论文，共223页。

提出了一种基于Haar小波变换的去噪算法，该方法基于最初开发的一种使用Tetrolet变换进行图像压缩的算法。Tetrolet变换是一种支持tetrominoes的自适应Haar小波变换，即通过连接四个大小相等的正方形而形成的形状。该算法对加性高斯白噪声（AWGN）图像进行Haar小波变换，并假设采用通用硬阈值，将峰值信噪比（PSNR）测量的去噪性能提高了1 - 2.5dB。该算法是基于局部的，在图像的每个4x4块上独立工作。与其他已发表的基于Haar小波变换的方法相比，它的性能也相当好（PSNR可以达到2dB的提升）。算法的局部性和Haar小波变换计算的简单性使得该算法非常适合于高效的硬件实现。

A new denoising algorithm based on the Haarwavelet transform is proposed. The methodology is based on an algorithminitially developed for image compression using the Tetrolet transform. TheTetrolet transform is an adaptive Haar wavelet transform whose support istetrominoes, that is, shapes made by connecting four equal sized squares. Theproposed algorithm improves denoising performance measured in peaksignal-to-noise ratio (PSNR) by 1-2.5 dB over the Haar wavelet transform forimages corrupted by additive white Gaussian noise (AWGN) assuming universalhard thresholding. The algorithm is local and works independently on each 4x4block of the image. It performs equally well when compared with other publishedHaar wavelet transform-based methods (achieves up to 2 dB better PSNR). Thelocal nature of the algorithm and the simplicity of the Haar wavelet transformcomputations make the proposed algorithm well suited for efficient hardwareimplementation.

1 引言

2 文献回顾

3 小波的作用

4 基于降噪的Tetrolet变换

5 性能分析

6 总结与结论

附录A Tetrominoe形状

附录B Matlab源代码

附录C 缩略语

更多精彩文章请关注公众号：