Easy2Hard: Learning to Solve the Intractables From a Synthetic Dataset for Structure-Preserving Image Smoothing

Feng, Yidan; Deng, Sen; Yan, Xuefeng; Yang, Xin; Wei, Mingqiang; Liu, Ligang

doi:10.1109/tnnls.2021.3084473

Cited by 26 publications

(14 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many supervised learning-based filters attempt to simulate the smoothing behaviors of the traditional filters, such as the deep edge-aware filter, 34 context aggregation network, 10 and trainable guided filter 11 . Other supervised learning-based filters strive to learn from the ground truth derived from manual labeling, 13 or synthetic data 12 . The neural network can also be trained in an unsupervised manner, 14 which can be considered to be learning an algorithm to solve self-supervised optimization problems.…”

Section: Related Workmentioning

confidence: 99%

“…11 Other supervised learning-based filters strive to learn from the ground truth derived from manual labeling, 13 or synthetic data. 12 The neural network can also be trained in an unsupervised manner, 14 which can be considered to be learning an algorithm to solve selfsupervised optimization problems.…”

Section: Deep Learning-based Filtersmentioning

confidence: 99%

“…The recent booming developments of deep learning techniques facilitate the edge-preserving filters, with image smoothing being achieved by training deep neural networks in either supervised 10 – 13 or unsupervised 14 manners. The deep learning techniques provide a general framework for various image filters.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Soft clustering based on high- and low-level features for image smoothing

Yang

Zeng

et al. 2023

J. Electron. Imag.

View full text Add to dashboard Cite

.Edge-preserving image smoothing is essential for computational imaging. Traditional filters are based on low-level image gradients; however, the edges defined on gradients do not align with the contours of human perception, thus suffering from certain limitations. We propose an image filter based on a soft clustering model that combines high-level semantics (derived from instance segmentation) with low-level features (intensities). The proposed filter works by first performing soft clustering on the input image according to the high- and low-level features to derive the affinity matrices, which are then fused for image smoothing. Experiment results indicate the advantages of the proposed filter in a variety of applications, including image smoothing, flash/non-flash fusion, detail enhancing, image dehazing, and depth upsampling. Furthermore, the proposed filter is efficient, and it takes 1.75 s to process a color image with 1 megapixel on a modern desktop.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Deep Learning-based Filtersmentioning

confidence: 99%

See 1 more Smart Citation

Soft clustering based on high- and low-level features for image smoothing

Yang

Zeng

et al. 2023

J. Electron. Imag.

View full text Add to dashboard Cite

show abstract

“…Most deep learning-based filters are based on supervised learning. 14,15,31,32 The smoothing quality is typically inferior to the filters they try to approximate. On the other hand, the optimization model of the global filters can be leveraged as the loss function to train the neural networks in an unsupervised manner.…”

Section: Related Workmentioning

confidence: 99%

“…13 The deep learning-based filters train neural networks for the purpose of image smoothing, such as the residual network 14 or joint edge and structure network. 15 The local filters are mostly efficient (or there exist fast approximations). However, the local nature limits their smoothing quality.…”

Section: Introductionmentioning

confidence: 99%

L0 image smoothing via iterating truncated L1 gradient regularization

et al. 2022

View full text Add to dashboard Cite

Edge-preserving image smoothing is a fundamental tool in computational photography and graphics. It aims to suppress insignificant details while maintaining salient structures. The classical L 0 filter provides an elegant framework for a variety of applications. However, it inclines to sharpen the salient edges, thus suffering from gradient reversals and color deviations. We propose a solution toward the objective of optimizing summed squared error regularized by the L 0 -norm of the gradients. The proposed solution explores an iterative strategy, where each iteration is an optimization problem based on the truncated L 1 regularization, which can be solved efficiently with the combination of alternating direction method of multipliers and Fourier domain optimization. Leveraging L 1 -norm limits the aggressive modifications of gradients during the iterative process, which alleviates various artifacts in classical L 0 filter. Experiment results indicate that the proposed method achieves superior performance in various applications, including texture removal, detail enhancement, HDR tone mapping, and compression artifact removal. The proposed filter can be conveniently implemented on GPU. It takes our filter 0.51 s to process a 720P color image on a NVIDIA GTX 1080 GPU.

show abstract