Layer Decomposition Learning Based on Gaussian Convolution Model and Residual Deblurring for Inverse Halftoning
Abstract
:1. Introduction
1.1. Image Decomposition in Deep Learning Frameworks
1.2. Residual Layer Design for Residual Learning
1.3. Residual Learning Problems for Inverse Halftoning
1.4. Progressively Residual Learning Problems for Inverse Halftoning
1.5. Contributions
- First, to design the base layer, a new statistical distribution of the image difference between a blurred continuous-tone image and a blurred halftoned image with a Gaussian filter with a narrow output range is shown. Based on this observation, the base layer is reconstructed using a new GCM-based residual subnetwork that predicts the difference between the blurred continuous-tone image and the blurred halftoned image; this method differs completely from the existing PRL [23,25], which uses an initial restored image from a DCNN for base layer generation.
- Second, the detail layer is generated based on structure-aware residual learning that predicts the difference image between the predicted base layer and the original image. To more effectively enhance image structures such as edges and textures, an image structure map predictor, which was introduced in a previous study [24], is incorporated into the residual detail layer learning, resulting in structure-enhancing learning. In addition, the predicted base layer is the low-pass-filtered version of the original image. Therefore, the proposed residual detail learning should be used to deblur the base layer, i.e., to remove the blurring of the base layer. This implies that the deblurring strategy is adopted in the proposed residual detail learning, unlike the existing PRL.
- Third, it is demonstrated that SALDL can be used to recover high-quality images from the predicted base layers whose quality is poor in terms of edge and texture representation. However, the existing PRL [23,25] cannot yield satisfactory results from the same base layers. This reveals that the existing PRL is not suitable for low-quality base layers. By contrast, the proposed structure-aware residual learning method is more effective for describing image structures. To our best knowledge, this is the first study that performed the abovementioned comparison, and the experimental results confirmed the feasibility of the proposed SALDL as a new PRL for inverse halftoning that surpasses state-of-the-art methods such as PRL, U-net, and DCNN.
2. Proposed SALDL Based on GCM
2.1. Motivations
2.2. Residual Layer Design for Baser Layer Generation
2.3. GCM-Based Residual Subnetwork for Baser Layer Generation
2.4. Detail Layer Design
2.5. Direct Deblurring Approach
2.6. Proposed Layer Decomposition Learning
3. Experimental Results
3.1. Training Data Collection
3.2. Networking Training
3.3. Visual Quality Evaluation
4. Conclusions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Donghui, L.; Takuma, K.; Takahiko, H.; Midori, T.; Kaku, S. Texture-aware error diffusion algorithm for multi-level digital halftoning. J. Imaging Sci. Technol. 2020, 64, 50410-1–50410-9. [Google Scholar]
- Wang, Y.; Huang, H.; Wang, C.; He, T.; Wang, J.; Nguyen, M.H. GIF2Video: Color dequantization and temporal interpolation of GIF images. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Long Beach, CA, USA, 15–20 June 2019; pp. 1419–1428. [Google Scholar]
- Do, H.C.; Cho, B.G.; Chien, S.I.; Tae, H.S. Improvement of low gray-level linearity using perceived luminance of human visual system in PDP-TV. IEEE Trans. Consum. Electron. 2005, 51, 204–209. [Google Scholar]
- Kao, W.-C.; Liu, C.-H.; Liou, S.-C.; Tsai, J.-C.; Hou, G.-H. Towards video display on electronic papers. J. Display Technol. 2016, 12, 129–135. [Google Scholar] [CrossRef]
- Son, C.-H.; Choo, H. Watermark detection of clustered halftoned images via learned dictionary. Signal Process. 2014, 102, 77–84. [Google Scholar] [CrossRef]
- Lieberman, D.J.; Allebach, J.P. A dual interpretation for direct binary search and its implications for tone reproduction and texture quality. IEEE Trans. Image Process. 2000, 9, 1950–1963. [Google Scholar] [CrossRef]
- Son, C.H. Inverse halftoning based on sparse representation. Opt. Lett. 2012, 37, 2352–2354. [Google Scholar] [CrossRef] [PubMed]
- Guo, J.-M.; Prasetyo, H.; Wong, K. Halftoning-based block truncation coding image restoration. J. Vis. Commun. Image Represent. 2016, 35, 193–197. [Google Scholar] [CrossRef]
- Guo, J.-M. Watermarking in dithered halftone images with embeddable cells selection and inverse halftoning. Signal Process. 2008, 88, 1496–1510. [Google Scholar] [CrossRef]
- Son, C.-H.; Lee, K.; Choo, H. Inverse color to black-and-white halftone conversion via dictionary learning and color mapping. Inf. Sci. 2015, 299, 1–19. [Google Scholar] [CrossRef]
- Kopf, J.; Lischinski, D. Digital reconstruction of halftoned color comics. ACM Trans. Graph. 2012, 31, 140. [Google Scholar] [CrossRef]
- Remez, T.; Litany, O.; Bronstein, A. A picture is worth a billion bits: Real-time image reconstruction from dense binary threshold pixels. In Proceedings of the IEEE International Conference on Computational Photography, Evanston, IL, USA, 13–15 May 2016; pp. 1–9. [Google Scholar]
- Zhang, E.; Zhang, Y.; Duan, J. Color inverse halftoning method with the correlation of multi-color components based on extreme learning machine. Appl. Sci. 2019, 9, 841. [Google Scholar] [CrossRef] [Green Version]
- Kite, T.D.; Damera-Venkata, N.; Evans, B.L.; Bovik, A.C. A fast, high-quality inverse halftoning algorithm for error diffused halftones. IEEE Trans. Image Process. 2000, 9, 1583–1592. [Google Scholar] [CrossRef]
- Stevenson, R. Inverse halftoning via MAP estimation. IEEE Trans. Image Process. 1997, 6, 574–583. [Google Scholar] [CrossRef] [PubMed]
- Foi, A.; Katkovnik, V.; Egiazarian, K.; Astola, J. Inverse halftoning based on the anisotropic LPA-ICI deconvolution. In Proceedings of the International TICSP Workshop on Spectral Methods and Multirate Signal Processing, Vienna, Austria, 11–12 September 2004; pp. 49–56. [Google Scholar]
- Son, C.-H.; Choo, H. Iterative inverse halftoning based on texture-enhancing deconvolution and error-compensating feedback. Signal Process. 2013, 93, 1126–1140. [Google Scholar] [CrossRef]
- Freitasa, P.G.; Fariasb, M.C.Q.; Araújo, A.P.F. Enhancing inverse halftoning via coupled dictionary training. Signal Process. Image Commun. 2016, 49, 1–8. [Google Scholar] [CrossRef]
- Son, C.-H.; Choo, H. Local learned dictionaries optimized to edge orientation for inverse halftoning. IEEE Trans. Image Process. 2014, 23, 2542–2556. [Google Scholar]
- Zhang, Y.; Zhang, E.; Chen, W.; Chen, Y.; Duan, J. Sparsity-based inverse halftoning via semi-coupled multi-dictionary learning and structural clustering. Eng. Appl. Artif. Intell. 2018, 72, 43–53. [Google Scholar] [CrossRef]
- Jimenez, F.P.; Miyatake, M.N.; Medina, K.T.; Perez, G.S.; Meana, H.P. An inverse halftoning algorithms based on neural networks and atomic functions. IEEE Latin Am. Trans. 2017, 15, 488–495. [Google Scholar] [CrossRef]
- Hou, X.; Qiu, G. Image companding and inverse halftoning using deep convolutional neural networks. arXiv 2017, arXiv:1707.00116. [Google Scholar]
- Xia, M.; Wong, T.-T. Deep inverse halftoning via progressively residual learning. In Proceedings of the Asian Conference on Computer Vision, Perth, Australia, 2–6 December 2018; pp. 523–539. [Google Scholar]
- Son, C.-H. Inverse halftoning through structure-aware deep convolutional neural networks. Signal Process. 2020, 173, 1–7. [Google Scholar] [CrossRef] [Green Version]
- Yuan, J.; Pan, C.; Zheng, Y.; Zhu, X.; Qin, Z.; Xiao, Y. Gradient-guided residual learning for inverse halftoning and image expanding. IEEE Access 2020, 8, 50995–51007. [Google Scholar] [CrossRef]
- Kang, L.-W.; Lin, C.-W.; Fu, Y.-H. Automatic single-image-based rain steaks removal via image decomposition. IEEE Trans. Image Process. 2012, 21, 1742–1755. [Google Scholar] [CrossRef] [PubMed]
- Lim, J.; Heo, M.; Lee, C.; Kim, C.-S. Contrast enhancement of noisy low-light images based on structure-texture-noise decomposition. J. Vis. Commun. Image Represent. 2017, 45, 107–121. [Google Scholar] [CrossRef]
- Son, C.-H.; Zhang, X.-P. Layer-based approach for image pair fusion. IEEE Trans. Image Process. 2016, 25, 2866–2881. [Google Scholar] [CrossRef] [PubMed]
- Starck, J.-L.; Fadili, J.; Murtagh, F. The undecimated wavelet decomposition and its reconstruction. IEEE Trans. Image Process. 2007, 16, 297–309. [Google Scholar] [CrossRef] [Green Version]
- Paris, S.; Hasinoff, S.W.; Kautz, J. Local laplacian filters: Edge-aware image processing with a Laplacian pyramid. Commun. ACM 2015, 53, 81–91. [Google Scholar] [CrossRef]
- Stark, J.L.; Elad, M.; Donoho, D.L. Image decomposition via the combination of sparse representation and a variational approach. IEEE Trans. Image Process. 2005, 14, 2675–2681. [Google Scholar] [CrossRef] [Green Version]
- Li, Y.; Tan, R.T.; Guo, X.; Lu, J.; Brown, M.S. Single image rain steak decomposition using layer priors. IEEE Trans. Image Process. 2017, 26, 3874–3885. [Google Scholar] [CrossRef]
- Tomasi, C.; Manduchi, R. Bilateral filtering for gray and color Images. In Proceedings of the IEEE International Conference on Computer Vision, Bombay, India, 4–7 January 1998; pp. 839–846. [Google Scholar]
- Li, S.; Kang, X.; Hu, J. Image fusion with guided filtering. IEEE Trans. Image Process. 2013, 27, 2864–2875. [Google Scholar]
- Ronneberger, O.; Fischer, P.; Brox, T. U-Net: Convolutional networks for biomedical image segmentation. arXiv 2015, arXiv:1505.04597. [Google Scholar]
- Lai, W.; Huang, J.; Ahuja, N.; Yang, M. Deep laplacian pyramid networks for fast and accurate super-resolution. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 624–632. [Google Scholar]
- Zhang, K.; Zuo, W.; Chen, Y.; Meng, D.; Zhang, L. Beyond a gaussian denoiser: Residual learning of deep CNN for image denoising. IEEE Trans. Image Process. 2017, 26, 3142–3155. [Google Scholar] [CrossRef] [Green Version]
- Fu, X.; Huang, J.; Zeng, D.; Huang, Y.; Ding, X.; Paisley, J. Removing rain from single images via a deep detail network. In Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 1–9. [Google Scholar]
- Hradiš, M.; Kotera, J.; Zemčík, P.; Šroubek, F. Convolutional neural networks for direct text deblurring. In Proceedings of the British Machine Vision Conference, Swansea, UK, 7–10 September 2015; pp. 6.1–6.13. [Google Scholar]
- Vedaldi, A.; Lenc, K. Matconvnet: Convolutional neural networks for matlab. In Proceedings of the ACM international conference on Multimedia, Brisbane, Australia, 26–30 October 2015; pp. 689–692. [Google Scholar]
- Wang, Z.; Bovik, A.C.; Sheikh, H.R.; Simoncelli, E.P. Image quality assessment: From error visibility to structure similarity. IEEE Trans. Image Process. 2004, 13, 600–612. [Google Scholar] [CrossRef] [Green Version]
- Kwon, J.-H.; Son, C.-H.; Cho, Y.-H.; Ha, Y.-H. Text-enhanced error diffusion using multiplicative parameters and error scaling factor. J. Imaging Sci. Technol. 2006, 50, 437–447. [Google Scholar] [CrossRef]
- Ruder, S. An overview of gradient descent optimization algorithms. arXiv 2016, arXiv:1609.04747. [Google Scholar]
Layers | Input Layer | Last Convolution Layer | Convolution Block (Number) | |
---|---|---|---|---|
Subnetworks | ||||
GCM-based residual subnetwork | ||||
IRS | (16) | |||
ISMP including IRS | (22) | |||
SARDS | (16) |
Methods | Proposed Method | U-Net [35] | DCNN [37] | DDN [39] | PRL [23,25] | |||||
---|---|---|---|---|---|---|---|---|---|---|
Test Images | PSNR | SSIM | PSNR | SSIM | PSNR | SSIM | PSNR | SSIM | PSNR | SSIM |
1 | 25.943 | 0.835 | 25.563 | 0.815 | 25.181 | 0.808 | 24.747 | 0.783 | 25.639 | 0.818 |
2 | 25.916 | 0.913 | 25.590 | 0.904 | 25.395 | 0.900 | 25.139 | 0.894 | 25.632 | 0.905 |
3 | 26.013 | 0.878 | 25.247 | 0.857 | 24.810 | 0.846 | 24.308 | 0.826 | 25.379 | 0.859 |
4 | 29.951 | 0.901 | 29.262 | 0.873 | 28.608 | 0.854 | 28.997 | 0.864 | 28.843 | 0.866 |
5 | 31.974 | 0.981 | 31.901 | 0.979 | 31.818 | 0.979 | 31.064 | 0.978 | 31.488 | 0.979 |
6 | 26.373 | 0.909 | 25.820 | 0.899 | 25.370 | 0.890 | 24.974 | 0.88 | 25.814 | 0.896 |
7 | 31.601 | 0.981 | 31.248 | 0.979 | 31.084 | 0.979 | 30.522 | 0.977 | 31.069 | 0.979 |
8 | 28.659 | 0.969 | 27.992 | 0.966 | 27.275 | 0.959 | 26.698 | 0.953 | 27.823 | 0.963 |
9 | 31.145 | 0.953 | 30.539 | 0.948 | 30.237 | 0.949 | 29.517 | 0.933 | 30.449 | 0.942 |
10 | 30.281 | 0.939 | 29.601 | 0.930 | 29.214 | 0.928 | 28.721 | 0.914 | 29.581 | 0.929 |
11 | 24.853 | 0.859 | 24.098 | 0.832 | 23.738 | 0.828 | 23.388 | 0.805 | 24.258 | 0.839 |
12 | 25.654 | 0.816 | 24.718 | 0.751 | 24.441 | 0.739 | 24.274 | 0.741 | 24.904 | 0.771 |
13 | 33.381 | 0.966 | 33.302 | 0.964 | 33.282 | 0.964 | 32.426 | 0.959 | 32.777 | 0.961 |
14 | 29.901 | 0.846 | 29.631 | 0.840 | 29.753 | 0.832 | 29.253 | 0.822 | 29.645 | 0.833 |
15 | 27.119 | 0.904 | 26.878 | 0.897 | 26.755 | 0.894 | 26.422 | 0.89 | 26.841 | 0.898 |
AVG. | 28.584 | 0.910 | 28.093 | 0.896 | 27.797 | 0.890 | 27.363 | 0.881 | 28.009 | 0.896 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Son, C.-H. Layer Decomposition Learning Based on Gaussian Convolution Model and Residual Deblurring for Inverse Halftoning. Appl. Sci. 2021, 11, 7006. https://doi.org/10.3390/app11157006
Son C-H. Layer Decomposition Learning Based on Gaussian Convolution Model and Residual Deblurring for Inverse Halftoning. Applied Sciences. 2021; 11(15):7006. https://doi.org/10.3390/app11157006
Chicago/Turabian StyleSon, Chang-Hwan. 2021. "Layer Decomposition Learning Based on Gaussian Convolution Model and Residual Deblurring for Inverse Halftoning" Applied Sciences 11, no. 15: 7006. https://doi.org/10.3390/app11157006
APA StyleSon, C.-H. (2021). Layer Decomposition Learning Based on Gaussian Convolution Model and Residual Deblurring for Inverse Halftoning. Applied Sciences, 11(15), 7006. https://doi.org/10.3390/app11157006