Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Nonlocal Adaptive Direction-Guided Structure Tensor Total Variation For Image Recovery

Ezgi Demircan-Tureyen, Mustafa E. Kamasak

TL;DR

This paper proposes a method that employs anisotropic Gaussian kernels to estimate directional features to be later used by the proposed model, and proves that the entire two-stage framework achieves better results than the NLSTV model and two other competing local models, in terms of visual and quantitative evaluation.

Abstract

A common strategy in variational image recovery is utilizing the nonlocal self-similarity (NSS) property, when designing energy functionals. One such contribution is nonlocal structure tensor total variation (NLSTV), which lies at the core of this study. This paper is concerned with boosting the NLSTV regularization term through the use of directional priors. More specifically, NLSTV is leveraged so that, at each image point, it gains more sensitivity in the direction that is presumed to have the minimum local variation. The actual difficulty here is capturing this directional information from the corrupted image. In this regard, we propose a method that employs anisotropic Gaussian kernels to estimate directional features to be later used by our proposed model. The experiments validate that our entire two-stage framework achieves better results than the NLSTV model and two other competing local models, in terms of visual and quantitative evaluation.

Nonlocal Adaptive Direction-Guided Structure Tensor Total Variation For Image Recovery

TL;DR

This paper proposes a method that employs anisotropic Gaussian kernels to estimate directional features to be later used by the proposed model, and proves that the entire two-stage framework achieves better results than the NLSTV model and two other competing local models, in terms of visual and quantitative evaluation.

Abstract

A common strategy in variational image recovery is utilizing the nonlocal self-similarity (NSS) property, when designing energy functionals. One such contribution is nonlocal structure tensor total variation (NLSTV), which lies at the core of this study. This paper is concerned with boosting the NLSTV regularization term through the use of directional priors. More specifically, NLSTV is leveraged so that, at each image point, it gains more sensitivity in the direction that is presumed to have the minimum local variation. The actual difficulty here is capturing this directional information from the corrupted image. In this regard, we propose a method that employs anisotropic Gaussian kernels to estimate directional features to be later used by our proposed model. The experiments validate that our entire two-stage framework achieves better results than the NLSTV model and two other competing local models, in terms of visual and quantitative evaluation.

Paper Structure

This paper contains 16 sections, 19 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. Method
  4. NLADSTV Regularizer
  5. Parameter Estimation
  6. Structure Tensor Eigendecomposition
  7. Postprocessing Initial Orientation Map
  8. Filtering with Anisotropic Gaussian Kernels
  9. Nonlocal Structure Tensor Eigendecomposition
  10. The Final Directional Parameters
  11. Overall Algorithm
  12. Experimental Results
  13. Image Denoising
  14. Image Deblurring
  15. Discussion
  16. ...and 1 more sections

Figures (4)

  • Figure 1: The flowchart of our DPE procedure for the estimation of directional parameters.
  • Figure 2: Thumbnails of the grayscale and color images used in the experiments. From left to right and top to bottom: Chateau, Dog, Mural, Workers, Buildings, Deer, Arch, Man, Monarch, and Structure.
  • Figure 3: The detail patches cropped from restored versions of the noisy Structure and Arch images. Structure image is degraded the noise level of $\sigma_\eta = 0.1$. Arch image is degraded by the noise level of $\sigma_\eta = 0.15$.
  • Figure 4: The detail patches cropped from restored versions of the blurred and noisy Buildings and Workers images. Buildings image is degraded by Gaussian blur and noise level of BSNR = 20 dBs. Workers image is degraded by motion blur and BSNR = 20 dBs.