Image Retrieval

本文综述了近年来利用深度学习技术进行图像检索的研究进展,涵盖了从高效特征提取到语义保持哈希等多种方法,并介绍了相关开源代码及数据集。

摘要生成于 C知道 ,由 DeepSeek-R1 满血版支持, 前往体验 >

Papers

1. Learning High-level Image Representation for Image Retrieval via Multi-Task DNN using Clickthrough Data

2.Neural Codes for Image Retrieval

3.Efficient On-the-fly Category Retrieval using ConvNets and GPUs

4.Deep Learning of Binary Hash Codes for Fast Image Retrieval

5.Learning visual similarity for product design with convolutional neural networks

6.Deep Semantic Ranking Based Hashing for Multi-Label Image Retrieval

7.Exploiting Local Features from Deep Networks for Image Retrieval

8.Supervised Learning of Semantics-Preserving Hashing via Deep Neural Networks for Large-Scale Image Search

9.Cross-domain Image Retrieval with a Dual Attribute-aware Ranking Network

10.Aggregating Deep Convolutional Features for Image Retrieval

11.Feature Learning based Deep Supervised Hashing with Pairwise Labels

12.Particular object retrieval with integral max-pooling of CNN activations

Group Invariant Deep Representations for Image Instance Retrieval

13.Where to Buy It: Matching Street Clothing Photos in Online Shops

14.Deep Image Retrieval: Learning global representations for image search

15.Bags of Local Convolutional Features for Scalable Instance Search

16.Faster R-CNN Features for Instance Search

17.Learning Compact Binary Descriptors with Unsupervised Deep Neural Networks

18.Deep Relative Distance Learning: Tell the Difference Between Similar Vehicles

19.DeepFashion: Powering Robust Clothes Recognition and Retrieval with Rich Annotations

20.CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples

21.SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval

22.Deep Semantic-Preserving and Ranking-Based Hashing for Image Retrieval

23.SIFT Meets CNN: A Decade Survey of Instance Retrieval

24.Deep Hashing: A Joint Approach for Image Signature Learning

25.End-to-end Learning of Deep Visual Representations for Image Retrieval

### 基于内容的图像检索技术及其实现方法 #### 定义与概述 基于内容的图像检索(Content-Based Image Retrieval, CBIR)是一种通过分析图像的内容特征来查找相似图片的技术。这种技术不依赖于人工标注或元数据,而是利用计算机视觉算法自动提取并匹配图像中的颜色、纹理、形状等低级视觉特征[^1]。 #### 关键组件和技术 CBIR系统通常由以下几个主要部分组成: - **特征提取**:此过程涉及从输入图像中抽取有意义的信息片段作为描述符。常见的特征类型包括但不限于直方图、边缘检测器响应以及局部不变量描述子如SIFT和SURF。 - **索引结构构建**:为了提高查询效率,在大规模数据库环境下建立有效的索引机制至关重要。常用的方法有树形结构、哈希表以及其他近似最近邻搜索策略。 - **相似度计算**:当接收到新的查询请求时,需将待查对象同已存入系统的样本集逐一比较其间的距离度量值;常用的衡量标准除了欧氏距离外还有余弦相似性和结构相似性指数(SSIM)[^2]。 #### 实现案例 以Riya为例,这是一个开源搜索引擎项目,它不仅支持传统的文本关键词匹配方式,还特别加入了针对人物肖像及商品照片等内容类型的专门处理模块——即融合了先进的面部识别技术和物品分类模型来进行更精准的结果筛选。 ```python import cv2 from skimage.metrics import structural_similarity as ssim def compare_images(imageA, imageB): # Convert images to grayscale grayA = cv2.cvtColor(imageA, cv2.COLOR_BGR2GRAY) grayB = cv2.cvtColor(imageB, cv2.COLOR_BGR2GRAY) # Compute SSIM between two images score, diff = ssim(grayA, grayB, full=True) return score ``` 上述代码展示了如何使用Python库`skimage`中的函数`structural_similarity()`来量化两张灰度化后的图像之间的结构性差异程度,返回一个介于[-1,+1]区间内的数值表示它们之间的一致性水平,其中越接近+1意味着更加相像。
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值