论文部分内容阅读
提出了一种基于非分样ridgelet标架的图像噪声滤除(UDRIFDA)的新算法。ridgelet标架的特点是:基函数不可分离变量且具有很强的方向性,能够实现对沿直线奇性的有效描述。离散非分样ridgelet标架是通过离散Radon变换切片上的一维非分样小波变换标架来实现的。由于非分样小波变换具有位移不变性,能够很好地刻画多尺度下一维信号的局部特征,基于一维非分样小波变换的软阈值去噪算法能够有效地降低一维信号急剧变化处所产生的震荡现象,故基于非分样ridgelet标架的图像滤噪算法能够大大降低恢复图像上的伪影,有效的克服了文献[1]中分样ridgelet标架滤噪算法(DRITDA)的缺陷。数值实验表明新算法较DRITDA和2D-DWT算法更能提高恢复图像的信噪比和视觉质量。
A new algorithm for image noise filtering (UDRIFDA) based on nonspecifed ridgelet frame is proposed. Ridgelet frame is characterized by: inseparable basis function variables and has a strong direction, to achieve an effective description along the straight line singularity. The discrete nonsubscribe ridgelet frame is realized by a one-dimensional non-decimating wavelet transform frame on a discrete Radon transform slice. Due to the invariance of the nonsampled wavelet transform, it can well characterize the local features of one-dimensional signals under multi-scale. The soft-thresholding algorithm based on one-dimensional non-decimating wavelet transform can effectively reduce the sharp changes of one-dimensional signals Therefore, the image filtering algorithm based on non-segmented ridgelet frame can greatly reduce the artifacts on the restored image and effectively overcome the defect of the ridgelet frame filter algorithm (DRITDA) in [1] . Numerical experiments show that the new algorithm can improve the signal-noise ratio and visual quality of the restored image better than the DRITDA and 2D-DWT algorithms.