霍文华,
苏明月,
付灏
1.燕山大学信息科学与工程学院 秦皇岛 066000
2.河北省信息传输与信号处理重点实验室 秦皇岛 066000
基金项目:国家自然科学基金(62072394)，河北省自然科学基金(F2021203019)
详细信息
作者简介:顾广华：男，1979年生，博士，教授，研究方向为图像检索、图像分类和图像识别
霍文华：女，1995年生，硕士生，研究方向为图像检索和深度哈希
苏明月：女，1996年生，硕士生，研究方向为图像检索和深度哈希
付灏：男，1996年生，硕士生，研究方向为跨模态检索
通讯作者:顾广华 guguanghua@ysu.edu.cn
中图分类号:TN911.73; TP391.4计量
文章访问数:83
HTML全文浏览量:44
PDF下载量:19
被引次数:0
出版历程
收稿日期:2020-11-23
录用日期:2021-11-05
修回日期:2021-10-25
网络出版日期:2021-11-09
刊出日期:2021-12-21
Asymmetric Supervised Deep Discrete Hashing Based Image Retrieval
Guanghua GU,,Wenhua HUO,
Mingyue SU,
Hao FU
1. School of Information Science and Engineering, Yanshan Engineering University, Qinhuangdao 066000, China
2. Hebei Key Laboratory of Information Transmission and Signal Processing, Qinhuangdao 066000, China
Funds:The National Natural Science Foundation of China (62072394), The Natural Science Foundation of Hebei Province (F2021203019)
摘要
摘要:哈希广泛应用于图像检索任务。针对现有深度监督哈希方法的局限性，该文提出了一种新的非对称监督深度离散哈希(ASDDH)方法来保持不同类别之间的语义结构，同时生成二进制码。首先利用深度网络提取图像特征，根据图像的语义标签来揭示每对图像之间的相似性。为了增强二进制码之间的相似性，并保证多标签语义保持，该文设计了一种非对称哈希方法，并利用多标签二进制码映射，使哈希码具有多标签语义信息。此外，引入二进制码的位平衡性对每个位进行平衡，鼓励所有训练样本中的–1和+1的数目近似。在两个常用数据集上的实验结果表明，该方法在图像检索方面的性能优于其他方法。
关键词:图像检索/
监督哈希/
语义保持/
深度学习
Abstract:Hashing is widely used for image retrieval tasks. In view of the limitations of existing deep supervised hashing methods, a new Asymmetric Supervised Deep Discrete Hashing (ASDDH) method is proposed to maintain the semantic structure between different categories and generate binary codes. Firstly, a deep network is used to extract image features and reveal the similarity between each pair of images according to their semantic labels. To enhance the similarity between binary codes and ensure the retention of multi-label semantics, this paper designs an asymmetric hashing method that utilizes a multi-label binary code mapping to make the hash codes have multi-label semantic information. In addition, the bit balance of the binary code is introduced to balance each bit, which encourages the number of -1 and +1 to be approximately similar among all training samples. Experimental results on two benchmark datasets show that the proposed method is superior to other methods in image retrieval.
Key words:Image retrieval/
Supervised hashing/
Semantic preservation/
Deep learning
PDF全文下载地址:
https://jeit.ac.cn/article/exportPdf?id=9a0a112d-f80c-4195-aa88-78dbfc67fba0
