To address the limitations of current digital halftoning algorithms, such as slow processing speed and suboptimal halftoning effects, a data-driven halftoning framework is proposed. By introducing the Gumbel-Softmax reparameterization strategy, the non-differentiability issue caused by discrete halftone selection is resolved, enabling unbiased gradient estimation during network backpropagation. To further enhance the halftoning effects, a novel blue noise loss function is designed to optimize the distribution of halftone dots. Additionally, a Patch-wise Confidence Aggregation module is introduced to incorporate spatial correlations between pixels, allowing the network to focus more on pixel interactions during training. Based on these strategies, a label-free, self-supervised, differentiable halftoning framework is constructed by optimizing the expected value of the halftone quality metric. Experimental results demonstrate that the proposed method, without requiring image labels, can generate high-quality halftone images and maintain high processing speed and low parameter complexity, effectively preserving local structural information and texture details. Moreover, this framework can be flexibly extended to multi-level halftoning to accommodate the requirements of multi-level printheads.
在半色调图像质量评估中,现有度量标准如均方误差和结构相似性指数虽然能够量化图像间差异,但它们往往受限于对像素级精确度的片面追求,忽略了像素间的相互作用对整体感知质量的影响.这些指标通过人类视觉系统模型对目标图像和参考图像进行预处理,以模拟人眼对图像细节的敏感度,继而生成反映图像差异的误差图并对其进行平均计算得到标量度量结果.像素的表现不仅仅依赖于其自身的值,还会受到其所在局部窗口内其他像素的影响,这一作用范围由HVS滤波器所限定的窗口大小决定.然而,仅通过广义平均池(generalized mean pooling)直接优化平均精度,将每个网络参数的梯度简化为所有像素处梯度的平均值,该处理方式将所有像素对的置信度成本等同对待,未能充分考虑到半色调图像中像素点排列模式对生成图像质量的关键影响[26-27].
ULICHNEYR A. Void-and-cluster method for dither array generation[C]//Human Vision,Visual Processing,and Digital Display IV. San Jose, CA. SPIE, 1993: 332-343.
[2]
MAOY F, ABELLOL, SARKARU, et al. 4-row serpentine tone dependent fast error diffusion[C]//2018 25th IEEE International Conference on Image Processing (ICIP). Athens, Greece. IEEE,2018: 3973-3977.
[3]
HUX Y .Simple gradient-based error-diffusion method[J].Journal of Electronic Imaging,2016,25(4):043029.
[4]
OSTROMOUKHOVV.A simple and efficient error-diffusion algorithm[C]//Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques. ACM,2001: 567-572.
[5]
FRANKT, LIUJ Y,GAT S,et al .A machine learning approach to design of aperiodic,clustered-dot halftone screens via direct binary search[J].IEEE Transactions on Image Processing,2022,31: 5498-5512.
[6]
LIAOJ R .Theoretical bounds of direct binary search halftoning[J]. IEEE Transactions on Image Processing,2015,24(11):3478-3487.
[7]
UFUK AGARA, ALLEBACHJ P .Model-based color halftoning using direct binary search[J]. IEEE Transactions on Image Processing, 2005, 14(12): 1945-1959.
WANGX H, LIUL L, CHENH,et al .A digital halftone algorithm based on the dynamic error diffusion coefficient[J].Packaging Engineering,2017,38(13):199-203.(in Chinese)
[10]
FUNGY H, CHANY H .Tone-dependent noise model for high-quality halftones[J].Journal of Electronic Imaging,2013,22(2):023004.
[11]
SHAOL H, ZHANGE H, LIM .An efficient convolutional neural network model combined with attention mechanism for inverse halftoning[J].Electronics,2021,10(13):1574.
[12]
JIANGH T, XIONGD L, JIANGX W,et al .Halftoning with multi-agent deep reinforcement learning[C]//2022 IEEE International Conference on Image Processing (ICIP). Bordeaux,France. IEEE,2022:641-645.
[13]
LAUC K, XIAM H, WONGT T. Taming reversible halftoning via predictive luminance[J]. IEEE Transactions on Visualization and Computer Graphics, 2024, 30(8): 4841-4852.
[14]
GUOJ M, SANKARASRINIVASANS .H-GAN:deep learning model for halftoning and its reconstruction[C]//2020 IEEE International Conference on Consumer Electronics (ICCE). Las Vegas,NV,USA. IEEE,2020:1-2.
[15]
CHOIB, ALLEBACHJ P. Mimicking DBS halftoning via a deep learning approach[J]. Electronic Imaging,2022,34(15):158-1-158-7.
[16]
LAUD L, ARCEG R. Modern digital halftoning[M]. Boca Raton:CRC Press,2018.
[17]
TITSIASM K, LÁZARO-GREDILLAM. Local expectation gradients for black box variational inference[J]. Advances in Neural Information Processing Systems, 2015: 1-9.
[18]
XIAM H, HUW B, LIUX T,et al .Deep halftoning with reversible binary pattern[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal,QC,Canada. IEEE,2021:13980-13989.
[19]
JIANGH T, XIONGD L, JIANGX W,et al .Efficient halftoning via deep reinforcement learning[J].IEEE Transactions on Image Processing,2023,32:5494-5508.
[20]
HEK M, ZHANGX Y, RENS Q,et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas,NV,USA. IEEE,2016:770-778.
LIS S, LIUH R, GANY D, et al. Image dehazing network based on residual dense block and attention mechanism[J].Journal of Hunan University (Natural Sciences), 2021, 48(6): 112-118.(in Chinese)
[23]
Unsupervised multi-object segmentation using attention and soft-argmax[C]//2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). Waikoloa, HI, USA. IEEE, 2023: 3266-3275.
[24]
HUIJBENI A M, KOOLW, PAULUSM B,et al .A review of the gumbel-max trick and its extensions for discrete stochasticity in machine learning[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(2): 1353-1371.
[25]
KIMS H, ALLEBACHJ P .Impact of HVS models on model-based halftoning[J]. IEEE Transactions on Image Processing,2002, 11(3): 258-269.
[26]
ULICHNEYR A .Dithering with blue noise[J]. Proceedings of the IEEE,1988,76(1):56-79.
[27]
ITOUAP, BEGHDADIA, VIARIS DE LESEGNOP .Objective perceptual evaluation of halftoning using image quality metrics[C]//10th International Conference on Information Science,Signal Processing and their Applications (ISSPA 2010). Kuala Lumpur,Malaysia. IEEE, 2010: 456-459.
[28]
FURUTAR, INOUEN, YAMASAKIT. PixelRL:fully convolutional network with reinforcement learning for image processing[J]. IEEE Transactions on Multimedia, 2020,22(7):1704-1719.
[29]
VASSILOK, HEATWOLEC, TAHAT,et al .Multi-step reinforcement learning for single image super-resolution[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Seattle,WA,USA. IEEE,2020: 2160-2168.
[30]
ABEDINIF, GOORANS .Structure-aware color halftoning with adaptive sharpness control[J]. Journal of Imaging Science and Technology, 2022, 66(6): 1-11.
基金资助
国家重点研发专项计划(2022YFE0112400)
National Key Research and Development Special Plan(2022YFE0112400)
国家自然科学基金资助项目(21706096)
National Natural ScienceFoundation of China(21706096)
“科技兴蒙”行动重点专项(NMKJXM202210)
The “Technology Empowering Inner Mongolia” Key Special Initiative(NMKJXM202210)