Sketch based image retrieval based on abstract-level transform and convolutional neural networks(融合抽象层级变换和卷积神经网络的手绘图像检索方法)

LIUYujie(刘玉杰); PANGYunping(庞芸萍); LIZongmin(李宗民); LIHua(李华)

doi:10.3785/j.issn.1008-9497.2016.06.005

Zhejiang Daxue xuebao. Lixue ban (Nov 2016)

Sketch based image retrieval based on abstract-level transform and convolutional neural networks(融合抽象层级变换和卷积神经网络的手绘图像检索方法)

LIUYujie(刘玉杰),
PANGYunping(庞芸萍),
LIZongmin(李宗民),
LIHua(李华)

Affiliations

LIUYujie(刘玉杰): ORCiD; 1.College of Computer & Communication Engineering, China University of Petroleum, Qingdao 266580, Shandong Province, China( 1.中国石油大学计算机与通信工程学院，山东　青岛　266580)
PANGYunping(庞芸萍): 1.College of Computer & Communication Engineering, China University of Petroleum, Qingdao 266580, Shandong Province, China( 1.中国石油大学计算机与通信工程学院，山东　青岛　266580)
LIZongmin(李宗民): ORCiD; 1.College of Computer & Communication Engineering, China University of Petroleum, Qingdao 266580, Shandong Province, China( 1.中国石油大学计算机与通信工程学院，山东　青岛　266580)
LIHua(李华): 2.Key Laboratory of Intelligent Information Processing, Institute of Computing Technology Chinese Academy of Sciences, Beijing 100190, China( 2.中国科学院计算技术研究所智能信息处理重点实验室，北京　100190)

DOI: https://doi.org/10.3785/j.issn.1008-9497.2016.06.005
Journal volume & issue: Vol. 43, no. 6
pp. 657 – 663

Abstract

Read online

针对人工设计的描述子(HOG、SIFT等）在基于手绘的图像检索(Sketch Based Image Retrieval，SBIR)领域的局限性，提出了一种融合抽象层级变换和卷积神经网络构建联合深度特征描述子的手绘图像检索方法.首先，提取常规图像的边缘概率图，在此基础上进行不同抽象层级的图像变换，将抽象层级变换图像输入到深度神经网络并提取不同隐层的输出向量，最后，联合不同隐层的输出向量作为手绘图像检索的特征描述子（即联合深度特征描述子）.在Flickr15k数据库上对本方法进行了实验验证，结果表明：融合抽象层级变换和联合深度特征描述子的检索效果相较HOG、SIFT等传统方法有显著提高.本方法从图像预处理和特征描述子构建2个方面，对SBIR问题进行了改进，具有更高的准确率.

Published in Zhejiang Daxue xuebao. Lixue ban

ISSN: 1008-9497 (Print)
Publisher: Zhejiang University Press
Country of publisher: China
LCC subjects: Science: Mathematics: Instruments and machines: Electronic computers. Computer science; Science: Physics
Website: https://www.zjujournals.com/sci/EN/1008-9497/home.shtml

About the journal

Abstract

Keywords