Denoising of digital images using modified recursive cycle-spinning and wavelet shrinkage

The denoising and accurate reconstruction of signals with fine details are challenging problems in signal processing. Traditional methods employ discrete wavelet transform-based techniques to address this issue. One such method, known as cycle spinning, estimates the original signal by averaging multiple denoised versions obtained from shifted and thresholded representations of the noisy input. In this study, we propose a modified recursive cycle spinning algorithm that enhances this approach by introducing a scaled-down threshold during the wavelet shrinkage step. The method involves computing the linear average of reconstructions derived from wavelet transforms of various shifted sequences of the noisy signal. This modification significantly improved computational efficiency while preserving the important signal characteristics. The effectiveness of the proposed method is demonstrated through a comparative analysis with existing techniques, including the standard recursive cycle spinning method, non-local means method, and dual-tree complex wavelet transform approach.