Nonlocal Gradient Sparsity Regularization for Image Restoration

Liu, Hangfan; Xiong, Ruiqin; Zhang, Xinfeng; Zhang, Yongbing; Ma, Siwei; Gao, Wen

doi:10.1109/tcsvt.2016.2556498

Cited by 66 publications

(25 citation statements)

References 62 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using the non-local mean theory, Dong et al [22] propose the non-local centralized sparse representation model to obtain the estimation of the sparse coding coefficients of the original image. By taking advantage of the non-local similarity of natural images, Liu et al [23] formulate the sparsity of the image gradient with pixel-wise content-adaptive distributions to reflect the non-stationary nature of image statistics. Daniel and Weiss [24] propose the Expected Patch Log Likelihood (EPLL) method by using a Gaussian mixture model (GMM) prior, which trains clean image patches to regularize degraded image patches.…”

Section: A Related Workmentioning

confidence: 99%

“…Here, D denotes the directional difference operator, D h j and D v j denote the horizontal and vertical directions respectively. Inspired by [23], the gradient difference between the target patch and similar patches is regularized by the total variation that is a typical L 1 -norm and has better performance in handling the directionality of the image. It can ensure the accuracy of prediction to avoid the disadvantage of the traditional total variation, which simply makes the regularization on the patch itself as zero.…”

Section: Patch-based Nonlocal Gradient Priormentioning

confidence: 99%

“…In image denoising, we compare the proposed method with several state-of-the-art methods, DnCNN [3], FID-HL [6], NLTV [20] and NGS [23]. The typical example on the image Cameraman is shown in Figure 5, Figure 5(a) is the noisy image degraded by the Gaussian noise with the standard deviation σ = 20.…”

Section: ) Image Denosingmentioning

confidence: 99%

“…In image inpainting, the date ratio r varies from 0.2 to 0.8. The comparison methods including TwIST [13], SALSA [16],TVAL3 [5] and NGS [23] are all state-of-theart methods and the recovered results on image Barbara are shown in Figure 6. Figure 6(a) is the masked image with the data ratio r = 0.4.…”

Section: ) Image Inpaintingmentioning

confidence: 99%

See 3 more Smart Citations

Patch-Based Nonlocal Adaptive Gradient Regularization for Image Restoration in Sensor Networks

Shi

2019

IEEE Access

View full text Add to dashboard Cite

In this paper, we propose a novel patch-based adaptive nonlocal gradient regularization method for image restoration in sensor networks. It formulates the hyper-Laplacian distribution to regularize the global gradient distribution. The patch-based nonlocal gradient prior is utilized to regularize the nonlocal self-similarity of image gradients. Firstly, the L 0-norm smoothing scheme is used innovatively as the preprocessing step to preserve strong edges, which are critical to improve the accuracy of clustering the similar image patches. Then, adaptive weights for each patches are developed from a set of clustered nonlocal self-similarity patches by learning the the expectation and variance for sparse gradient distribution at each pixel. Comparing with several recent state-of-the-art methods, experimental results show that the proposed method has better performance in alleviating block effects and preserving image details. INDEX TERMS L 0-norm regularization, image restoration, hyper-Laplacian, image priors.

show abstract

Section: A Related Workmentioning

confidence: 99%

Section: Patch-based Nonlocal Gradient Priormentioning

confidence: 99%

Section: ) Image Denosingmentioning

confidence: 99%

Section: ) Image Inpaintingmentioning

confidence: 99%

See 2 more Smart Citations

Patch-Based Nonlocal Adaptive Gradient Regularization for Image Restoration in Sensor Networks

Shi

2019

IEEE Access

View full text Add to dashboard Cite

show abstract

“…China; Tel: 0938-8366657; E-mail: xdlq@163.com image quality. The observed image is usually the degradation image under factors of the noise, fuzzy, sports translation [3]. The classic image restoration process is based on prior knowledge to build an image degradation model, and then a variety of the degradation processing methods are adopted to make the image quality improved.…”

Section: Introductionmentioning

confidence: 99%

Microscopic Images Restoration of Chinese Herbal Medicine Using the TV Model

Liu¹,

Zhang²,

Yang³

et al. 2015

TOAUTOCJ

View full text Add to dashboard Cite

Chinese herbal medicine (CHM) samples defiled, extraction of microscopic images, collection site light not equal and images collected uneven illumination point which can make the acquisition of CHM microscopic images subject to different unfavorable factors. Those often make microscopic images of CHM become blurred or introduce noise. In this paper, some restoration methods for the microscopic images of CHM using the total variation (TV) model are put forward. Firstly, the microscopic images of CHM are blurred utilizing the two dimensional different Gaussian convolution kernels and different Gauss white noise are added, and then the damaged images are restored by using the improved TV model. The TV restoration results are compared with the results using Wiener filtering, median filtering method etc. The theoretical analysis and experimental results show that the novel method is effectively able to restore image, and maintain the image detail information.

show abstract