基于Sobel滤波器的图像风格转换

doi:10.19678/j.issn.1000-3428.0059961

摘要/Abstract

摘要： 基于迭代的图像风格转换是将图像的内容与风格信息进行分离然后重组，在图像重组的时候没有考虑内容图像的结构问题，导致生成的图像存在线条扭曲。针对这一问题，提出了一种基于边缘检测的算法来约束图像重组时的信息。边缘检测使用 Sobel 滤波器，用卷积神经网络在内容图像和生成图像相同的卷积层进行边缘信息的提取，损失函数选择均方误差，将边缘损失、内容损失和风格损失的加权代数和作为神经网络的总损失。实验结果表明，通过加入边缘检测的方法进行图像风格转换，有效的抑制了图像的线条扭曲，降低了图像的噪声，使得生成了更高质量的图像。

Abstract: The iterative image style conversion is to separate and reorganize the content and style information of the image. When the image is reorganized, the structure of the content image is not considered, resulting in distortion of lines in the generated image. To solve this problem, an algorithm based on edge detection is proposed to constrain the information during image reorganization. Edge detection uses sobel filter, and convolutional neural network is used to extract edge information in the same convolutional layer of content image and generated image. The loss function selects the mean square error, and the weighted algebraic sum of edge loss, content loss and style loss The total loss of the neural network. The experimental results show that the image style conversion by adding the edge detection method effectively suppresses the distortion of the image lines, reduces the image noise, and produces a higher quality image.

陈志鹏 , 郑文秀 , 黄琼丹. 基于Sobel滤波器的图像风格转换[J]. 计算机工程, doi: 10.19678/j.issn.1000-3428.0059961.

Chen Zhipeng, Zheng Wenxiu, Huang QiongDan. Image style conversion based on sobel filter[J]. Computer Engineering, doi: 10.19678/j.issn.1000-3428.0059961.

参考文献

[1] 张景祥.王士同.邓赵红.等.融合异构特征的子空间迁移学习算法[J]. 自动化学报. 2014.(2):236-246. ZHANG Jingxiang. Wang Sh itong. DENG Zhaohong et al. A subspace transfer lea rning algorithm integrating heterogeneous features[J]. Acta automatica sinica 2014,40(2):236-246
[2] Portilla J, Simoncelli E P. A Parametric Texture Mod el Based on Joint Statistics of Complex Wavelet Coef ficients[J]. International Journal of Computer Vision, 2 000(40):49-70.
[3] Semmo A , Limberger D.，Kyprianidis J.E , Dollner J. Image Stylization by Oil Paint Filtering Using Col or Palettes[J]. Workshop on Computational Aesthetics. Potsdam, 2015:149158.
[4] Shih Y.C, Paris S, Barnes C, Freeman W.T, Durand F. Style Transfer for Headshot Portraits[J]. ACM Transa ctions on Graphics, 2014(33):1-14.
[5] LI X.T，LIU S.F，Kautz J，YANG M.H. Learning Li near Transformations for Fast Arbitrary Style Transfer [J]，arXiv: 1808.04537.2018
[6] CHEN D.D，YUAN L，LIAO J，YU N.H.，HUA G. StyleBank:An Explicit Representation for Neural Image Style Transfer.[J]. arXiv:1506:02753. 2017
[7] Gatys L A, Ecker A S, Bethge M. A Neural Algor-it hm of Artistic Style[DB]. eprint ar-Xiv:1508.06576, 2 015.
[8] Gatys L A, Ecker A S, Bethge M. Image Style Trans fer Using Convolutional Neural Networks[C]//2016 IE EE Conference on Computer Vision and Pattern Reco gnition(CVPR).IEEE, 2016:2414-2423.
[9] 郑茗化, 白本督等人. 基于局部均方差的神经网络图像风格转换[J]. 现代电子技术, 2019.14.033. Zheng Minghua BAI Bendu et al.Neural Network Image Style Conversion Based o n Local Mean Square Error[J]. Modern ElectronicsTec hniqu. 2019.14.033
[10] Risser E, Wilmot P, Barnes C. Stable And Controllabl e Neural Texture Synthesis And Style Transfer Using Histogram Losses[J]. ComputerScience, 2017, arXiv: 1701.08893.
[11] Li C, Wand M. Combining Markov RandomFields An d Convolution Neural Networks For Image Synthesis [C]. Computer Vision andPattern Recognition, IEEE, 2016:2479-248.
[12] BERGER G, MEMISEVIC R. Incorporating long rang e consistency in CNN-based texture generation[J]. Co mputer Science,arXiv:1606,2016.
[13] LI Y, LIU M Y, LI X,et al. A closed-form solution t o photo-realistic image stylization[EB/OL].[2019-07-23]. https://arxiv.org/pdf/1802.06474.pdf.
[14] Goodfellow I J, Pouget-Abadie J, Mirza M, et al. Ge nerative adversarial nets[C]//Proceedings of the 28th I nternational Conference on Neural Information Process ing Systems. Massachusetts: MIT Press,2014:2672-268 0.
[15] HUANG X, BELONGIE S. Arbitrary style transfer in real-time with adaptive instance normalization[J]. IEE E International Conference on Computer Vision, 2017: 1510-1519.
[16] 张航,曹瞻,基于局部均值与方差的图像中值滤波方法 [J].中南大学学报(自然科学版), 2013, 44(z2):381-38 4. ZHANG Hang, CAO Zhan. Image median filter m ethod based on local mean and variance[J]. Journal of Central South University(Science and technology), 20 13, 44(S2): 381-384.
[17] LIU Xiuwen, LIU Zhi, ZHOU Xiaofei,et all.Saliency- guided image style transfer[C]//International Conferenc e on Multimedia & Expo Workshops, 2019:66-71.
[18] ZHU Ting, LIU Shiguang. Detail-preserving arbitrary stylle transfer[C]//In Proceedings of the IEEE Internati onal Conference on Multimedia and Expo(ICME), 202 0:1-6.
[19] 谢志峰, 叶冠桦, 闫淑萁,等. 基于生成对抗网络的 H DR 图像风格迁移技术[J]. 上海大学学报(自然科学版), 2018, 24(4):524-534. XIE Zhifeng, YE Guanh ua, YAN Shuqi, et al HDR image style transfer techn ology based on generative confrontation network[J].Jo urnal of Shanghai University (Natural Science Edition), 2018, 24(4):524-534.
[20] Simonyan K, Zisserman A. Very Deep Con-volutional Networks for Large-Scale Image Recognition[J]. epri nt arXiv:1409.1556,2014.
[21] 高飞, 黄启宏, 程卫东等. 边缘检测算子的改进研究 [J]. 数字技术与应用. 2016(01) GAO Fei, HUAN G Qihong, CHENG Weidong. et al. Research on Impr ovement of EdgeDete , ction Operator[J]. Digital Tech nology and Application. 2016(01)
[22] 虎玲, 常霞, 纪峰. 图像边缘检测方法研究新进展[J]. 现代电子技术. 2018(23) HU ling, Chang xia ,JI Feng. New Progress in Research on Image Edge Detection[J]. Modern Electronics Technique. 2018(23)
[23] 苗英杰，崔琛. 基于 BFGS 拟牛顿法的观测矩阵优化算法[J]. 电子信息对抗技术. 2019.06.008 MIAO Yingjie, CUI Shen. Observation matrix optimization algorithm based on BF GS quasi-Newton method[J]. Electronic Information W arfare Technology. 2019.06.008