LightGCNet: Deep Gamut Compression Algorithm Based on Lightweight Convolutional Networks

doi:10.3778/j.issn.1673-9418.2402017

Abstract

Abstract: Gamut compression is a key technology for maintaining color information when converting from a wide to a narrow color gamut. Classical algorithms are fast but their results lack detail and do not consider the perceptual characteristics of the human eye, failing to meet the increasing demands for color quality. Iterative algorithms offer better results but are extremely time-consuming and not feasible for practical application. To address these issues, this paper proposes a lightweight deep gamut compression method, maintaining speed close to classical algorithms while approaching the effectiveness of iterative algorithms. Initially, this paper designs the LightGCNet model for gamut compression, featuring dual convolution layers composed of DSC and DW convolutions, drastically reducing parameters and computational complexity compared with conventional U-Net. Moreover, to further enhance the model prior knowledge, the approach refocuses multiple kernel channels in the DW convolutions obtained from pre-training to establish connections, with final weights comprising both refocused and pre-trained weights. Subsequently, a specialized loss function for gamut compression is designed for hue, luminance, and chroma. This function considers pixel-level loss, incorporates perceptual loss, and weights multiscale color attribute information, addressing detail loss in deep gamut compression. Finally, labels generated by iterative algorithms are used to train the network, combining learning of target gamut features and image information, achieving results comparable to iterative algorithms. Experimental results show that compared with the industry-standard SGCK algorithm, the iCID value is reduced by 17.08%, and the SSIM value is increased by 5.30%. Compared with the conventional U-Net model, the number of parameters in LightGCNet is reduced by 82.96%, and the number of multiply-accumulate operations is decreased dramatically from 8.5 GFLOPs to 2.2 GFLOPs, making the improved model more suitable for deployment on low-end devices. The model processes a single 512×512 image on CPU in just 0.208 s, achieving a 99.92% reduction in computation time compared with iterative algorithms.

Key words: gamut compression, lightweight, deep learning, color management

摘要： 色域压缩是大色域向小色域进行转换时保持色彩信息的关键技术。经典算法虽计算快速，但处理结果缺乏细节且没有考虑人眼的感知特性，难以满足人们对色彩品质不断提升的要求。迭代算法处理效果更佳，但极其耗时，无法投入实际应用。针对上述问题，提出了轻量化的深度色域压缩方法，该方法可在逼近迭代算法计算效果的同时保持接近经典算法的速度。为色域压缩算法设计了LightGCNet模型，该网络中的双层卷积由DSC与DW卷积组合而成，相比常规U-Net，参数量与计算复杂度急剧降低。为进一步提高模型先验性，将预训练得到的DW卷积中多个核通道进行重聚焦操作以建立联系，最终权重由重聚焦权重和预训练权重组成。针对色相、明度、彩度这三个色彩属性，设计了色域压缩专用的损失函数。该函数不仅考虑了像素级损失，而且融合了图像感知损失，并联合色彩属性的多尺度信息进行加权，解决了深度色域压缩过程中细节丢失的问题。基于迭代算法生成标签以训练网络，联合学习目标色域特征与图像信息，实现了与迭代算法相当的效果。实验结果表明：该方法与业界经典的SGCK算法相比，iCID值降低了17.08%，SSIM值提高了5.30%。相比常规U-Net模型，LightGCNet参数量减少了82.96%，乘加次数从8.5 GFLOPs大幅下降至2.2 GFLOPs，使改良模型更适于低端设备部署。该模型在CPU上处理单幅512×512图像仅需0.208 s，计算时间比迭代类算法减少99.92%。

关键词: 色域压缩, 轻量化, 深度学习, 色彩管理

YANG Chen, XU Hao, ZHU Jiawei, WU Qin, CHAI Zhilei. LightGCNet: Deep Gamut Compression Algorithm Based on Lightweight Convolutional Networks[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(1): 196-210.

杨晨, 徐昊, 朱佳伟, 吴秦, 柴志雷. LightGCNet：基于轻量化卷积网络的深度色域压缩算法[J]. 计算机科学与探索, 2025, 19(1): 196-210.

References

[1] MOROVIC J. Color gamut mapping[M]. Chichester: John Wiley & Sons, 2008.
[2] LE H M, JEONG T, ABDELHAMED A, et al. GamutNet: restoring wide-gamut colors for camera-captured images[C]//Proceedings of the 29th Color and Imaging Conference, Nov 1-4, 2021: 7-12.
[3] LE H M, PRICE B, COHEN S, et al. GamutMLP: a lightweight MLP for color loss recovery[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 18268-18277.
[4] CHOI B, ALLEBACH J P. Mimicking DBS halftoning via a deep learning approach[J]. Electronic Imaging, 2022, 34(15).
[5] XIA M, HU W, LIU X, et al. Deep halftoning with reversible binary pattern[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 14000-14009.
[6] KASSON J M. Color image gamut-mapping system with chroma enhancement at human-insensitive spatial frequencies: U.S. Patent 5450216[P]. 1995-09-12.
[7] FARUP I, HARDEBERG J Y, AMSRUD M. Enhancing the SGCK colour gamut mapping algorithm[C]//Proceedings of the 2nd European Conference on Colour in Graphics, Imaging, and Vision, Aachen, Apr 5-8, 2004: 520-524.
[8] XU L, ZHAO B, LUO M R. Colour gamut mapping between small and large colour gamuts: part I. gamut compression[J]. Optics Express, 2018, 26(9): 11481-11495.
[9] HAN H, KWAK Y. How Koreans understand colorfulness, chroma, vividness, and depth[C]//Proceedings of the 30th Color and Imaging Conference, Scottsdale, Nov 13-17, 2022: 246-249.
[10] LI H, ZHAO H, LIU M, et al. Research on spatial gamut mapping from source device to destination device[C]//Proceedings of the 2023 IEEE 18th Conference on Industrial Electronics and Applications. Piscataway: IEEE, 2023: 1909-1914.
[11] 刘菊华. 彩色打印机色彩特性化关键技术研究[D]. 武汉: 武汉大学, 2014.
LIU J H. Research on key technologies for color printer characterization[D]. Wuhan: Wuhan University, 2014.
[12] CHOI B, HU S, GUO R, et al. Developing a gamut mapping method for a novel inkjet printer[J]. Electronic Imaging, 2022, 34: 1-6.
[13] BALA R, DEQUEIROZ R, ESCHBACH R, et al. Gamut mapping to preserve spatial luminance variations[J]. Journal of Imaging Science and Technology, 2001, 45(5): 436-443.
[14] ZOLLIKER P, SIMON K. Adding local contrast to global gamut mapping algorithms[C]//Proceedings of the 3rd European Conference on Colour in Graphics, Imaging, and Vision, Leeds, Jun 19-22, 2006: 257-261.
[15] ZHU M, HARDEBERG J Y, WANG N, et al. Spatial gamut mapping based on guided filter[J]. Electronic Imaging, 2016(20): 1-4.
[16] ZHU M, ZHUN T. A spatial gamut mapping algorithm based on adaptive detail preservation[J]. Journal of Imaging Science and Technology, 2018, 62(1).
[17] 朱明, 焦会敏, 赵兴运, 等. 一种基于图像细节保持的空间色域映射算法框架[J]. 北京理工大学学报, 2020, 40(3): 290-297.
ZHU M, JIAO H M, ZHAO X Y, et al. A spatial gamut mapping framework based on detail-preservation[J]. Transactions of Beijing lnstitute of Technology, 2020, 40(3): 290-297.
[18] NAKAUCHI S, HATANAKA S, USUI S. Color gamut mapping based on a perceptual image difference measure[J]. Color Research & Application, 1999, 24(4): 280-291.
[19] KIMMEL R, SHAKED D, ELAD M, et al. Space-dependent color gamut mapping: a variational approach[J]. IEEE Transactions on Image Processing, 2005, 14(6): 796-803.
[20] BONNIER N, SCHMITT F, BRETTEL H, et al. Evaluation of spatial gamut mapping algorithms[C]//Proceedings of the 14th Color and Imaging Conference, Scottsdale, Nov 6-10, 2006: 56-61.
[21] ALSAM A, FARUP I. Spatial colour gamut mapping by orthogonal projection of gradients onto constant hue lines[C]//Proceedings of the 2012 International Symposium on Visual Computing. Berlin, Heidelberg: Springer, 2012: 556-565.
[22] WANG Y, ZENG P, LUO X. Color gamut mapping based on image fusion[C]//Proceedings of the 2008 International Conference on Computer Science and Software Engineering. Piscataway: IEEE, 2008: 801-805.
[23] PREISS J, FERNANDES F, URBAN P. Color-image quality assessment: from prediction to optimization[J]. IEEE Transactions on Image Processing, 2014, 23(3): 1366-1378.
[24] LISSNER I, URBAN P. Toward a unified color space for perception-based image processing[J]. IEEE Transactions on Image Processing, 2011, 21(3): 1153-1168.
[25] LIU S, LI S. Gamut mapping optimization algorithm based on gamut-mapped image measure (GMIM)[J]. Signal, Image and Video Processing, 2018, 12: 67-74.
[26] TAKEUCHI M, SAKAMOTO Y, YOKOYAMA R, et al. A gamut-extension method considering color information restoration using convolutional neural networks[C]//Proceedings of the 2019 IEEE International Conference on Image Processing. Piscataway: IEEE, 2019: 774-778.
[27] LI H, LIU Q, SUN B, et al. A novel gamut expansion method based on combined global-local mapping for sRGB-to-ProPhoto conversion[J]. IET Image Processing, 2023, 17(11): 3165-3176.
[28] CHOLLET F. Xception: deep learning with depthwise separable convolutions[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 1251-1258.
[29] HOWARD A G, ZHU M, CHEN B, et al. Mobilenets: efficient convolutional neural networks for mobile vision applications[EB/OL]. [2023-12-24]. https://arxiv.org/abs/1704. 04861.
[30] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]//Proceedings of the 18th International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, Oct 5-9, 2015. Cham: Springer, 2015: 234-241.
[31] CAI Z, DING X, SHEN Q, et al. RefConv: re-parameterized refocusing convolution for powerful ConvNets[EB/OL]. [2023-12-24]. https://arxiv.org/abs/2310.10563.
[32] ZHAO H, GALLO O, FROSIO I, et al. Loss functions for image restoration with neural networks[J]. IEEE Transactions on Computational Imaging, 2016, 3(1): 47-57.
[33] WANG Z, SIMONCELLI E P, BOVIK A C. Multiscale structural similarity for image quality assessment[C]//Proceedings of the 37th Asilomar Conference on Signals, Systems & Computers. Piscataway: IEEE, 2003: 1398-1402.
[34] REINHARD J, URBAN P. Perceptually optimizing color look-up tables[J]. IEEE Transactions on Image Processing, 2022, 32: 403-414.
[35] ZAMIR S W, ARORA A, KHAN S, et al. Restormer: efficient transformer for high-resolution image restoration[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 5728-5739.
[36] VALANARASU J M J, PATEL V M. UNeXt: MLP-based rapid medical image segmentation network[C]//Proceedings of the 2022 International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer, 2022: 23-33.
[37] YIN Y, HAN Z, JIAN M, et al. AMSUnet: a neural network using atrous multi-scale convolution for medical image segmentation[J]. Computers in Biology and Medicine, 2023, 162: 107120.
[38] MAAZ M, SHAKER A, CHOLAKKAL H, et al. EdgeNeXt: efficiently amalgamated CNN-Transformer architecture for mobile vision applications[C]//Proceedings of the 17th European Conference on Computer Vision. Cham: Springer, 2022: 3-20.
[39] LI Y, YUAN G, WEN Y, et al. Efficientformer: vision transformers at mobilenet speed[C]//Advances in Neural Information Processing Systems 35, New Orleans, Nov 28-Dec 9, 2022: 12934-12949.
[40] LIANG J, CAO J, SUN G, et al. SwinIR: image restoration using swin transformer[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 1833-1844.