Face Hallucination via Gradient Constrained Sparse Representation

Pei, Xiaobing; Guan, Yue; Cai, Pingping; Dong, Tao

doi:10.1109/access.2018.2795038

Cited by 10 publications

(9 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to effectively alleviate the problems above, Local patch-based face image recovery approaches have been proposed recently [16]- [17], [25]- [29], [31]- [35]. In these approaches, the face is divided into small patches, reconstructed separately, and then stitched into a complete face image.…”

Section: New Methods Have Beenmentioning

confidence: 99%

“…In our previous work [35], I-GCSR model is proposed ( Fig. 1), which exploits the gradient information to improve the face super-resolution result.…”

Section: New Methods Have Beenmentioning

confidence: 99%

“…where 0 ≤ λ ≤ 1 is the trade-off parameter, when λ = 0 equation (4) is reduced to weighted 2 -norm, when λ = 1 equation 4is reduced to weighted 1 -norm, that is the same as our last work [35].…”

Section: B Weighted Elastic Network Sparse Representation Modelmentioning

confidence: 97%

“…For the gradient of input image G X , we first enlarge the image to the desired size with the bicubic interpolation, and then calculate the gradient images. We adopt the strategy of patch super-resolution reconstruction [34], [35]. Due to the manifold consistency between LR and HR, we can use the weighted linear combination of HR face image patches in the training set to obtain the reconstructed HR output face image patch [34]- [35].…”

Section: Face Super-resolution Via Wensr Modelmentioning

confidence: 99%

See 3 more Smart Citations

Super-Resolution of Face Images Using Weighted Elastic Net Constrained Sparse Representation

2019

Self Cite

View full text Add to dashboard Cite

High-resolution (HR) face images are usually preferred in many computer vision tasks. However, low-resolution (LR) face images, which are often obtained in real scenarios, can be converted to a high resolution with the super-resolution techniques. In this paper, we propose the weighted elastic net constrained sparse representation (WENSR) super resolution method for face images. The method considers image gradient and weighted elastic net penalties. Due to the high similarity between human faces, it is not very suitable to only use 1-norm in the sparse representation model. The elastic net has a grouping effect and is more suitable for real-world data. The gradient is very important information in the image, we also use image gradient to enhance the final output. The tests of our method on both synthetic data and real-world data, such as FEI, CAS-PEAL-R1, and CMU+MIT face image dataset suggest a competitive performance gain in terms of peak signal to noise ratio (PSNR) and structural similarity (SSIM).

show abstract

Section: New Methods Have Beenmentioning

confidence: 99%

“…In our previous work [35], I-GCSR model is proposed ( Fig. 1), which exploits the gradient information to improve the face super-resolution result.…”

Section: New Methods Have Beenmentioning

confidence: 99%

Section: B Weighted Elastic Network Sparse Representation Modelmentioning

confidence: 97%

Section: Face Super-resolution Via Wensr Modelmentioning

confidence: 99%

See 2 more Smart Citations

Super-Resolution of Face Images Using Weighted Elastic Net Constrained Sparse Representation

2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…Based on LcR, recently, the low-rank and self-similarity priors are also introduced to regularize patch representation in [31], [32], [33]. In [34], Pei et al incorporated the gradient information of face image to further regularize the patch representation. In addition to face hallucination, the LcR algorithm has been also used to deal with pose and illumination problems in face hallucination and synthesis [9], [10], [35].…”

Section: Introductionmentioning

confidence: 99%

Context-Patch Face Hallucination Based on Thresholding Locality-Constrained Representation and Reproducing Learning

Jiang

Tang

et al. 2020

IEEE Trans. Cybern.

View full text Add to dashboard Cite

Face hallucination is a technique that reconstruct high-resolution (HR) faces from low-resolution (LR) faces, by using the prior knowledge learned from HR/LR face pairs. Most state-of-the-arts leverage position-patch prior knowledge of human face to estimate the optimal representation coefficients for each image patch. However, they focus only the position information and usually ignore the context information of image patch. In addition, when they are confronted with misalignment or the Small Sample Size (SSS) problem, the hallucination performance is very poor. To this end, this study incorporates the contextual information of image patch and proposes a powerful and efficient context-patch based face hallucination approach, namely Thresholding Locality-constrained Representation and Reproducing learning (TLcR-RL). Under the context-patch based framework, we advance a thresholding based representation method to enhance the reconstruction accuracy and reduce the computational complexity. To further improve the performance of the proposed algorithm, we propose a promotion strategy called reproducing learning. By adding the estimated HR face to the training set, which can simulates the case that the HR version of the input LR face is present in the training set, thus iteratively enhancing the final hallucination result. Experiments demonstrate that the proposed TLcR-RL method achieves a substantial increase in the hallucinated results, both subjectively and objectively. Additionally, the proposed framework is more robust to face misalignment and the SSS problem, and its hallucinated HR face is still very good when the LR test face is from the real-world. The MATLAB source code is available at https://github.com/junjun-jiang/TLcR-RL.

show abstract

Gaussian noise robust face hallucination via average filtering based data fidelity and locality regularization

Rajput

2022

Appl Intell

View full text Add to dashboard Cite

Face Hallucination via Gradient Constrained Sparse Representation

Cited by 10 publications

References 31 publications

Super-Resolution of Face Images Using Weighted Elastic Net Constrained Sparse Representation

Super-Resolution of Face Images Using Weighted Elastic Net Constrained Sparse Representation

Context-Patch Face Hallucination Based on Thresholding Locality-Constrained Representation and Reproducing Learning

Gaussian noise robust face hallucination via average filtering based data fidelity and locality regularization

Contact Info

Product

Resources

About