In this project , we presented a novel gradient histogram preservation (GHP) model for texture-enhanced image denoising, and further introduce two region-based GHP variants, i.e., B-GHP and S-GHP. A simple but theoretically solid model and the associated algorithm were presented to estimate the reference gradient histogram from the noisy image, and an efficient iterative histogram specification algorithm was developed to implement the GHP model. By pushing the gradient histogram of the denoised image toward the reference histogram, GHP achieves promising results in enhancing the texture structure while removing random noise. The experimental results demonstrated the effectiveness of GHP in texture enhanced image denoising. GHP leads to similar PSNR/SSIM measures to the state-of-the-art denoising methods such as SAPCABM3D, LSSC and NCSR; however, it leads to more natural and visually pleasant denoising results by better preserving the image texture areas. Most of the state-of-the-art denoising algorithms are based on the local sparsity and nonlocal selfsimilarity priors of natural images. Unlike them, the gradient histogram used in our GHP method is a kind of global prior, which is adaptively estimated from the given noisy image.
Zuo, Wangmeng, et al. “Gradient histogram estimation and preservation for texture enhanced image denoising” IEEE transactions on image processing 23.6 (2014): 2459-2472.