基于小波预测的超分辨率素描人脸合成方法  

Super-resolution sketch face synthesis via wavelet prediction

作　　者：王鑫玮 朱希安[2] 张本奎 杜康宁 郭亚男 WANG Xin-wei;ZHU Xi-an;ZHANG Ben-kui;DU Kang-ning;GUO Ya-nan(Key Laboratory of Ministry of Education for Optoelectronic Measurement Technology and Instrument,Beijing Information Science and Technology University,Beijing 100101,China;School of Information and Communication Engineering,Beijing Information Science and Technology University,Beijing 100101,China;Geography&Cyberspace Information Technology Research Department,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100080,China)

机构地区：[1]北京信息科技大学光电测试技术及仪器教育部重点实验室,北京100101 [2]北京信息科技大学信息与通信工程学院,北京100101 [3]中国科学院空天信息创新研究院地理与赛博空间信息技术研究部,北京100080

出　　处：《计算机工程与设计》2022年第5期1374-1381,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61671069、62001033);北京信息科技大学“勤信人才”培育计划基金项目(QXTCPA201902);北京市教委面上基金项目(KM202011232021)。

摘　　要：在使用人工智能算法等途径生成高质量的素描人脸图像时,若输入光学图像分辨率较低,生成的素描图像质量也会较差。在仅有低分辨率光学图像的条件下,提出一种基于小波预测的超分辨率素描人脸合成方法合成高质量的素描图像。在素描人脸合成网络的基础上引入超分辨率模块,通过对高分辨率图像的小波包分解系数进行预测,在端到端的框架下同时完成对图像的素描人脸图像合成以及超分辨率重建,提高合成图像的质量及分辨率。通过在CUHK学生数据集上与目前领先的超分辨率重建方法进行实验对比,该方法相较其它对比方法取得了更加优越的实验结果。When using artificial intelligence algorithm and other approaches to generate high-quality sketch face images,if the optical image resolution is low,the generated sketch image’s quality will also decrease.A super-resolution sketch face synthesis method based on wavelet prediction was proposed to improve the quality of sketches generated from low-resolution optical images.The super-resolution module was introduced for the sketch face synthesis network.By predicting the wavelet packet decomposition coefficients of the high-resolution image,the sketch face image synthesis and the super-resolution reconstruction of the image were completed simultaneously under the end-to-end framework,thereby improving the quality and resolution of the synthesized image.Experiments were implemented on the CUHK student data set and were compared with the current leading super-resolution reconstruction methods.Experimental results are better than other methods.

关 键 词：素描人脸合成 超分辨率重建 深度学习 高质量图像合成 小波包分解 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]