SepLUT: Separable Image-Adaptive Lookup Tables for Real-Time Image Enhancement

Yang, Canqian; Jin, Meiguang; Xu, Yi; Zhang, Rui; Chen, Ying; Liu, Huaida

doi:10.1007/978-3-031-19797-0_12

Cited by 11 publications

(9 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although DeepLPF achieved good results, its predicted differentiable filter shapes are lim- Example of the retouching results on the MIT-Adobe-5K evaluation dataset. For every two rows in order: input image, CURL [42], DeepUPE [41], DeepLPF [13], FRL [26], Exposure [10], E-GAN [4], RUAS [43], PixelRL [25], NeurOp [19], A3DLUT [16], SepLUT [17], AdaInt [18], ours, and ground truth. Zoom in to better see the details.…”

Section: B Experimental Resultsmentioning

confidence: 99%

“…Hui Zeng et al [16] proposed a method that uses a small neural network to adaptively predict the fusion weights of 3D LUTs for fast and real-time image enhancement. Canqian Yang et al [17] decomposed the color transform into two parts: color component-independent and color componentcorrelated, and used one-dimensional and three-dimensional LUTs to represent them respectively, thereby improving the efficiency and quality. They also designed a 3D LUT method [18] that adaptively learns the sampling interval, and proposed a new loss function to balance fidelity and visual quality.…”

Section: A Image Enhancementmentioning

confidence: 99%

“…In all the experiments, ↑ higher is better, ↓ lower is better, the best result is in bold form. 20.97 0.841 0.056 Exposure [10] 18.57 0.701 0.078 PixelRL [25] 13.60 0.630 0.227 FRL [26] 16.73 0.821 0.088 Ours 24.31 0.913 0.052 DPE [40] 22.15 0.850 0.075 E-GAN [4] 15.26 0.774 0.098 DeepUPE [41] 25.22 0.895 0.053 DeepLPF [13] 21.65 0.858 0.055 CURL [42] 24.23 0.865 0.053 RUAS [43] 17.88 0.856 0.086 NeurOp [19] 21.97 0.872 0.078 A3DLUT [16] 25.12 0.922 0.051 SepLUT [17] 25.21 0.920 0.052 AdaInt [18] 25.31 0.923 0.051…”

Section: ) Evaluation Metricsmentioning

confidence: 99%

See 2 more Smart Citations

Image Enhancement Using Adaptive Region- Guided Multi-Step Exposure Fusion Based on Reinforcement Learning

Xi¹,

Chen³

et al. 2023

IEEE Access

View full text Add to dashboard Cite

In photography, accurate exposure is key to taking high-quality photos, particularly images with uneven exposure levels. Global exposure operations are usually difficult to effectively strengthen the various regions of the image. To achieve a balanced exposure level in different regions of the image, we propose a novel algorithm inspired by the luminance masking frequently employed by professional photographers. We use reinforcement learning to adaptively generate guiding regions and adjusting parameters as a basis for multi-step exposure fusion, to enhance the overall quality of the image. Firstly, reinforcement learning is employed to automatically segment the single image to be enhanced into multiple sub-images, with corresponding appropriate adjusting parameters for each sub-image generated. Then, the input image is enhanced using the local adjusting parameters, yielding a set of images with varying enhancing degrees. Finally, these images are fused in an exposure process to obtain the final result. Experimental results show that our method not only generates intuitive and interpretable guiding regions, but also its performance is comparable to that of other contemporaneous methods.INDEX TERMS Deep reinforcement learning, Image enhancement, Multi-step decision.

show abstract

Section: B Experimental Resultsmentioning

confidence: 99%

Section: A Image Enhancementmentioning

confidence: 99%

Section: ) Evaluation Metricsmentioning

confidence: 99%

See 1 more Smart Citation

Image Enhancement Using Adaptive Region- Guided Multi-Step Exposure Fusion Based on Reinforcement Learning

Xi¹,

Chen³

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…More recently, deep learning techniques have been employed for estimating 3D LUTs and learning interpolation and sampling strategies within the LUT with the goal of color image enhancement [30,46,49]. For example, Zeng et al [50] proposed a method that learned the parameters of three 3D LUTs and an additional per-image adaption to blend between the LUTs.…”

Section: D Luts For Color Manipulationmentioning

confidence: 99%

“…They proposed a learned method to improve the classical trilinear interpolation between uniform sampling points in the LUT [29]. Also, Yang et al [49] proposed to learn separated component-correlated sub-transforms as 1D and 3D LUTs.…”

Section: D Luts For Color Manipulationmentioning

confidence: 99%