Unsupervised Texture Enhanced Image Denoising Using Gradient Histogram Estimation and Preservation

Removal of noise uses various natural image priors, including gradient based, sparse representationbased and nonlocal self similarity-based ones. The existing denoising algorithm tends to smooth the fine scale image textures, when removing noise, it degrades the image visual quality. To address this problem, a texture enhanced image denoising method is introduced. As a result, the gradient histogram of the denoised image, will be much closer with reference gradient histogram of the original image. Based on this, the gradient histogram preservation (GHP) algorithm is developed to enhance the texture structure while removing noise. Two region-based variants of GHP are proposed for the denoising of images consisting of region with different textures. An algorithm is also developed to effectively estimate the reference gradient histogram from the noisy observation of the unknown image by using unsupervised parameterization method. The experimental results demonstrate that the GHP algorithm can well preserve the texture appearance in the denoised images. To evaluate the performance of proposed algorithm, PSNR and SSIM values will be calculated and the same will be compared with existing denoising algorithms.