A region fusion based split Bregman method for TV Denoising algorithm

2021 
A novel robust image denoising method based Total Variation (TV) model called Region Fusion based Split Bregman (RFSB) method is proposed in this paper. First, the structural characteristics of the edge region and that of the smooth region of the noisy image are analyzed and then separated. Second, split Bregman method is used for calculating two baseline TV-based model, the TV model and TV-penalized least squares functional model, and results of the two models are fused by proposed Region Fusion method. Third, for the fusion denoising result contains a wealth of repetitive redundant information, fast non-local means (FNLM) is introduced for post-processing the denoising result. Our proposed method has three contributions: (i) the use of split Bregman to calculate two baseline TV-based model because of its highly minimization speed; (ii) the proposal of a region fusion method to fuse two baseline models; (iii) the application of the FNLM method as the post-processing. Gray images and color images with different intensities of noise are used to compare our proposed method with other existing classical TV-based method. Experimental results show that Our proposed method has outperformed other TV-based methods in denoising visual effect and objective evaluation for different styles of noised images. The time complexity of our proposed method is about 2% and 50% lower than that of TV and Zhang’s Total Variation (ZTV). Our proposed method shows lower time complexity and does not distort the image information compared with two deep learning algorithm.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    37
    References
    1
    Citations
    NaN
    KQI
    []