基于改进型正交有限脊波分析的自适应图像去噪

Adaptive image denoising based on modified orthonormal finite ridgelet transform

提出了一种改进型正交有限脊波分析的图像去噪算法。正交有限脊波分析处理信息时,有限域平面上的直线由于模运算的存在而引起信息Radon系数的卷绕,为此,对Radon系数周期振荡方向的信息采用离散余弦变换处理,其他方向信息采用小波分析处理,以减少卷绕对信息重建影响;同时,结合图像信息中噪声的特点,提出了一种去噪浮动阈值的选取方案。仿真结果表明,采用该算法实现的图像去噪,在抑制噪声的同时较有效地保留了信号的细节,去噪图像的PSNR值较OFR IT+W iener、2D-DWT等算法有所增加。

One kind of Modified Orthonormal Finite Ridgelet Transform （MOFRIT） was brought forward. The information Radon coefficient wrap-around of the line on a finite field was caused by the modular arithmetic so that the Radon coefficient information in the periodic oscillations direction was processed with discrete cosine transformation, and other direction information was processed with wavelet analysis in order to reduce the wrap-around of rebuilding information. At the same time, according to the characteristic of the noise, one kind of denoising threshold selection plan was proposed. Experimental results show that the details of the signal with above algorithm are more effectively retained and the denoising image PSNR value is higher than the value of OFRIT ＋ Wiener and 2D-DWT.