High-Dynamic-Range Image De-ghosting Fusion Method Based on Coherency Sensitive Hashing Patch-Match
-
摘要: 高动态范围(High dynamic range, HDR)图像成像技术的出现, 为解决由于采集设备动态范围不足而导致现有数字图像动态范围有限的问题提供了一条切实可行的思路.合成高动态范围图像的过程中因相机抖动或运动物体所造成的模糊和伪影问题, 可通过块匹配对多曝光图像序列进行去伪影融合加以解决.但对于具有复杂运动变化的真实场景, 现有的去伪影融合方法准确度和效率仍存在不足.为此, 本文结合相机响应函数和一致性敏感哈希提出了一种高动态图像去伪影融合方法.仿真结果表明, 该方法有效降低了计算复杂度, 具有较好的鲁棒性, 在有效去除伪影的同时提升了高动态范围图像质量.Abstract: The emergence of high dynamic range (HDR) image imaging technology provides a practical idea for solving the problem of limited dynamic range of the existing digital images which caused by the insu–cient dynamic range of the acquisition device. In the process of synthesizing high dynamic range images, the problem of blurring and artifacts caused by camera shake or moving objects can be solved by patchmatch in multi-exposure image sequences de-ghosting fusion. However, the accuracy and e–ciency of existing de-ghosting fusion methods are still poor for real scenes with complex motion changes. To this end, our paper presents a high dynamic image de-ghosting fusion method based on camera response function and coherency sensitive hashing. The simulation results show that the proposed method can efiectively reduce the computational complexity and has good robustness, and enhance the quality of high dynamic range image while efiectively removing artifacts.
-
Key words:
- High dynamic range (HDR) image /
- patchmatch /
- camera response function /
- coherency sensitive hashing /
- de-ghosting fusion
1) 本文责任编委 刘青山 -
图 2 WH变换投影向量
说明:其中s = [1], 白色表示其像素值为1, 黑色表示为-1, 且WH向量以递增的空间变化频率排序
Fig. 2 WH transform projection vector
图 4 基于相机响应函数和CSH的去伪影融合方法流程图
Fig. 4 Flow chart of de-ghosting fusion method based on CRF and CSH
图 8 对Forrest图像序列使用不同方法得到的相机响应函数
Fig. 8 The obtained CRF curve by using different methods in Forrest Sequence
图 9 Forrest图像序列去伪影融合方法的对比
Fig. 9 Comparison of different de-ghosting algorithms in Forrest sequence
图 10 Forrest图像序列去伪影融合方法的局部细节对比
Fig. 10 Local detail comparison of de-ghosting fusion methods in Forrest sequence
图 11 SculptureGarden图像序列去伪影融合方法的对比
Fig. 11 Comparison of different de-ghosting algorithms in SculptureGarden
图 12 SculptureGarden图像序列去伪影融合方法的局部细节对比
Fig. 12 Local detail comparison of de-ghosting fusion methods in SculptureGarden sequence
图 13 children_and_slide图像序列去伪影融合方法的对比
Fig. 13 Comparison of different de-ghosting algorithms in children_and_slide
图 14 children_and_slide图像序列去伪影融合方法的局部细节对比
Fig. 14 Local detail comparison of de-ghosting fusion methods in children_and_slide
图 15 fast_abrupt_motion图像序列去伪影融合方法的对比
Fig. 15 Comparison of different de-ghosting algorithms in fast_abrupt_motion
图 16 fast_abrupt_motion图像序列去伪影融合方法的对比
Fig. 16 Comparison of different de-ghosting algorithms in fast_abrupt_motion
表 1 根据观察1~4组合得到3种图像A中块a的候选块
Table 1 According to the observation 1 ~ 4, 3 kinds of candidates of block a in A is obtained
类型 定义 使用的观察类型 1 gB-1(gA(a)) 1和3 2 gB-1(gB(Right(Cand(Left(a))))) 3和4 3 Cand(gA-1(gA(a))) 2 
下载: 导出CSV
表 2 改进前后算法性能对比
Table 2 Performance comparison of Debevec$ ' $s and our algorithm
多组图像序列 图像分辨率 Debevec方法(s) 改进方法(s) 平均误差 Forrest $ 683 \times 1\, 024 \times 3 $ 0.555 0.368 $ 12.36\% $ Arch $ 1\, 024 \times 669 \times 3 $ 1.249 0.633 $ 7.68\% $ fast_abrupt_motion $ 1\, 080 \times 1\, 920 \times 3 $ 0.445 0.275 $ 11.44\% $
下载: 导出CSV
表 3 对于不同图像序列, 不同的去伪影方法评价指标
Table 3 For difierent image sequences, performe of difierent de-ghosting methods
图像序列 方法 平均亮度 对比度 信息熵 运行时间(s) Forrest RM 106.101 25.583 7.421 37.103 HU 132.799 27.603 7.546 122.547 PSSV 97.636 48.641 7.838 16.307 Ours 132.265 33.263 7.635 7.848 SculptureGarden RM 107.484 18.151 7.551 60.981 HU 122.785 21.187 7.325 182.470 PSSV 87.305 20.096 7.332 22.359 Ours 127.772 23.920 7.553 8.752 children_and_slide RM 90.813 11.943 7.491 86.038 HU 91.018 14.363 7.649 133.990 PSSV 85.190 15.499 7.717 21.235 Ours 93.342 14.949 7.668 8.194 fast_abrupt_motion RM 109.083 10.365 7.335 97.803 HU 122.893 13.337 7.662 152.608 PSSV 102.323 14.699 7.781 20.185 Ours 122.920 12.415 7.636 7.630
下载: 导出CSV
-
[1] Banterle F, Artusi A, Debattista K, et al. Advanced High Dynamic Range Imaging: Theory and Practice. Boca Raton, USA: CRC Press, 2011. [2] Reinhard E, Heidrich W, Debevec P, et al. High Dynamic Range Imaging: Acquisition, Display, and Image-based Lighting. San Francisco, USA: Morgan Kaufmann, 2010. [3] 陆许明, 朱雄泳, 李智文, 等.一种亮度可控与细节保持的高动态范围图像色调映射方法.自动化学报, 2015, 41(6): 1080-1092 doi: 10.16383/j.aas.2015.c130202Lu Xu-Ming, Zhu Xiong-Yong, Li Zhi-Wen, et al. A brightness-scaling and detail-preserving tone mapping method for high dynamic range images. Acta Automatica Sinica, 2015, 41(6): 1080-1092 doi: 10.16383/j.aas.2015.c130202 [4] Debevec P E, Malik J. Recovering high dynamic range radiance maps from photographs. In: Proceedings of the 24th Annual Conference on Computer Graphics and Interactive Techniques, New York, USA: ACM, 1997. 369-378 [5] Barnes C, Shechtman E, Finkelstein A, et al. PatchMatch:a randomized correspondence algorithm for structural image editing. ACM Transactions on Graphics-TOG, 2009, 28(3):1-11 http://d.old.wanfangdata.com.cn/Periodical/dbdxxb201605001 [6] Barnes C, Shechtman E, Goldman D B, et al. The generalized patchmatch correspondence algorithm. In: Proceedings of the Computer Vision-ECCV 2010, Berlin Heidelberg, Germany: Springer 2010. 29-43 [7] Sen P, Kalantari N K, Yaesoubi M, et al. Robust patchbased HDR reconstruction of dynamic scenes. ACM Trans. Graph., 2012, 31(6): 203 doi: 10.1145/2366145.2366222 [8] Jun Hu, Orazio Gallo, Kari Pulli, et al. HDR deghosting: how to deal with saturation? In: Proceedings of the 2013 IEEE Conference on Computer Vision and Pattern Recognition. Portland, OR, USA: IEEE, 2013. 1163-1170 [9] Korman S, Avidan S. Coherency sensitive hashing. In: Proceedings of the 2011 IEEE International Conference on Computer Vision. Barcelona, Spain: IEEE, 2011. 1607-1614 [10] Datar M, Immorlica N, Indyk P, et al. Locality-sensitive hashing scheme based on p-stable distributions. In: Proceedings of the 20th Annual Symposium on Computational Geometry. Brooklyn, New York, USA: ACM, 2004. 253-262 [11] Shen F M, Xu Y, Liu L, et al. Unsupervised deep hashing with similarity-adaptive and discrete optimization. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(12): 3034-3044 doi: 10.1109/TPAMI.2018.2789887 [12] Gu Zheng-Hui. The difierence method for the recovery of the camera response function. New Industrialization, 2013, 01: 15-22 [13] 方华猛, 易本顺, 甘良才, 等.高动态范围图像合成中相机响应函数的快速标定.光子学报, 2013, 06(4): 737-741 http://d.old.wanfangdata.com.cn/Periodical/gzxb201306020Fang Hua-Meng, Yi Ben-Shun, Gan Liang-Cai, et al. Fast calibration of camera response function in high dynamic range image synthesis. Journal of Photons, 2013, 06(4):737-741 http://d.old.wanfangdata.com.cn/Periodical/gzxb201306020 [14] Gionis A, Indyk P, Motwani R. Similarity search in high dimensions via hashing. In: Proceedings of the 25th VLDB Conference. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc., 1999. 518-529 [15] Hel-Or Y, Hel-Or H. Real-time pattern matching using projection kernels. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2005, 27(9): 1430-1445 doi: 10.1109/TPAMI.2005.184 [16] Ben-Artzi G, Hel-Or H, Hel-Or Y. The gray-code fllter kernels. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007, 29(3): 382-393 doi: 10.1109/TPAMI.2007.62 [17] Gallo O, Gelfand N, Chen W C, et al. Artifact-free high dynamic range imaging. In: Proceedings of the 2009 IEEE International Conference on Computational Photography. San Francisco, CA, USA: IEEE, 2009. 1-7 [18] Karaduzovic-Hadziabdic K, Telalovic J H, Mantiuk R. Evaluation of deghosting algorithms for multi-exposure high dynamic range imaging. In: Proceedings of the 29th Spring Conference on Computer Graphics (SCCG013). New York, NY, USA: ACM, 2013. 21-28 [19] Oh T H, Lee J Y, Tai Y W, et al. Robust high dynamic range imaging by rank minimization. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(6):1219-1232 doi: 10.1109/TPAMI.2014.2361338 [20] Oh T H, Kim H, Tai Y W, et al. Partial sum minimization of singular values in RPCA for low-level vision. In: Proceedings of the 2013 IEEE International Conference on Computer Vision. Sydney, NSW, Australia: IEEE, 2013. 145-152 [21] Oh T H, Tai Y W, Bazin J C, et al. Partial sum minimization of singular values in robust PCA: algorithm and applications. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(4): 744-758 doi: 10.1109/TPAMI.2015.2465956 -
计量
- 文章访问数: 1468
- HTML全文浏览量: 308
- PDF下载量: 144
- 被引次数: 0
