Adaptive bidirectional diffusion for image restoration

Abstract

A large number of applications in image processing and computer vision depend on image quality. In this paper, combining the forward diffusion with the backward diffusion by different weights, we present an adaptive bidirectional diffusion method for image denoising and deblurring simultaneously. Further, we introduce a gradient factor into the data fidelity term, which forms a spatially varying constraint and allows a better restoration of image edges and fine details. In order to obtain a stable solution, we develop a numerically stable scheme and give its theoretical analysis. Finally, we show advantages of this method in deblurring edges, denoising and restoring fine details of image compared with other related methods in experiments.

Bilateral signal variance estimation for wavelet-domain image denoising

“…The parameters used in the proposed approach are adaptive to noise statistics as σ d = σ n and σ r = 5 × σ n . In the first experiment, we evaluate the noise model formulated in (1). We calculate the noisy wavelet coefficients as defined in (1) and evaluate their distributions by comparing their histograms and the fitted Gaussian distributions.…”

“…In the first experiment, we evaluate the noise model formulated in (1). We calculate the noisy wavelet coefficients as defined in (1) and evaluate their distributions by comparing their histograms and the fitted Gaussian distributions. As seen in Figure 2, the Gaussian distribution is able to fit the histogram of the noisy wavelet coefficients very well.…”

“…Images are usually corrupted with noises during image acquisition and transmission [1,2]. Wavelet-based algorithms have proved effective for tackling the image denoising problem.…”