Noise level estimation is a long-standing problem in image processing. The challenge arises from the estimation being easily affected by texture information. We propose an innovative noise level estimation method via the kurtosis test, which is a normalized fourth-order moment. The proposed method consists of two stages: the first one is to determine the image patches with normality using the kurtosis test; the noise level is then estimated from these selected normal patches in the second stage, in which the average of the standard deviations is used as the final estimation. Experimental results show that the proposed method outperforms the state-of-the-art estimation techniques in terms of noise level estimation and guided denoising.
Parameter setting and information redundancy are essential issues for the non-local means (NLM) algorithm. This paper introduces a new factor based on the Hessian matrix to adapt the smoothing parameter. Then, a strategy is proposed to implement the NLM by representing patches in terms of features, which uses the 2D histogram and summed-area table. Compared with other methods, the metric for patch similarity in this paper is based on statistical features of patches instead of Euclidean distance. More importantly, not many predefined thresholds are needed. Experimental results show that the proposed algorithm obtains better visual quality and numerical results, especially for images with rich contents and high noise.
14 February 2020 in IEEE Computer Graphics and Applications
Achieving high visibility and high SNR (signal-to-noise ratio) from a single-shot image captured in low-light environments is an under-constrained problem. To cope with this issue, the intrinsic relationship between the image domain and the radiance domain is first established based on the human visual model, the atmospheric scattering model, and the camera imaging model, and the ideal exposure is derived. Using the illumination-reflection-noise prior, a new convex optimization by employed gradient constraint and Krisch operator is then presented to estimate the noise-reduced illumination and reflection components. A high SNR image in the optimal exposure is generated in radiance domain, which is finally inversely mapped to obtain a high SNR image in image domain. Experimental results in subjective and objective tests show that the proposed algorithm has a high SNR and pleasant perception in comparison with the state-of-the-art methods.
Patch level-based noise level estimation (NLE) is often inaccurate and inefficient because of the harsh criteria required to select a small number of homogeneous patches. In this paper, a fast image NLE method based on a global search for similar pixels is proposed to solve the above problem. Specifically, the mean square distance (MSD) is first expressed in the form of the standard deviation (std) and mean value of image patches. Afterward, the two values, std and mean, are calculated and stored in advance. Then, a 2D statistical histogram and summed area table are adopted to speed up the search for similar patches. Further, the most similar pixels are selected from similar patches to obtain an initial estimation. Finally, we correct the deviation of the initial estimation by re-injecting noise for secondary estimation. Experimental results show that the proposed method outperforms the state-of-the-art techniques in fast NLE and guided denoising.
Guided image filtering (GIF) is a popular edge-preserving smoothing technique, and the regularization parameter selection plays an important role in the performance of GIF. In this paper, we propose a new guided image filter based on Hessian matrix which consists of the second-order derivatives of an image. More specifically, a new structural measurement index is introduced by using the eigenvalues of the Hessian matrix first, which can distinguish the texture and flat regions of the image. Then the regularization parameter is adjusted based on this Hessian-based second-order structure measurement index, that is, a large regularization parameter is selected to improve the smoothness of the flat regions, while a small regularization parameter is set for the texture regions to preserve the image structure such as edges and corners. To further improve the quality of the filtered images, we also introduce a weighted averaging technique to the linear filter coefficients based on local variance. Experimental results show that the proposed Hessian-based weighted guided image filtering (HWGIF) method outperforms the state-of-the-art approaches in image processing applications such as edge-preserving denoising, detail enhancement, dehazing and HDR compression.