目录
基于内容的图像检索系统 1
主要任务： 1
详细功能描述： 2
摘要 2
一、系统设计 3
1.1 主要开发工具 3
1.2 实验方法与步骤 3
1.2.1 图像检索系统工作流程 3
1.2.2 图像检索系统框架 4
1.2.3 系统设计 6
二、所采用算法思想 9
2.1 图像预处理 9
2.2 mid 9
2.2.1 图像颜色特征 9
= 11
2.2.2 图像纹理特征 11
2.2.3 灰度共生矩阵法 12
2.2.4 形状不变矩法 14
= ∑ 14
2.2.5 边缘方向直方图法 15
其中 = { 17
1 或 = 0 17
min(, ) / ∑ 17
=1 17
=1 18
=1 18
=1 18
三、详细实现过程 18
3.1 图像特征数据库设计说明 18
3.2 Java 包与类的设计 20
四、实验结果分析 24
五、实训总结和心得 34
六、参考文献 34
基于内容的图像检索系统
主要任务：
设计与实现一个基于内容的图像检索系统，包括前端显示和后台服务器，后台服务器采用数据库存放图像特征数据，前台输入一个图片检索系统中对应的图片并进行显示
详细功能描述：
实现基于颜色信息的图像检索，可通过颜色直方图、颜色矩、颜色一致性矢量等方法来实现。
实现基于纹理特征的图像检索，可从四个方面进行：统计法、结构法、模型法、频谱法。
实现基于形状特征的图像检索，可分别从图像的边缘信息和区域信息来实现。
实现基于综合信息的图像检索。
预期成果或目标：
完成基于三种图像特征(颜色、纹理、形状)5 种相似度量算法的图像检索系统
指导老师评语：
指导教师签字：
摘要
随着数字化信息应用领域的不断深入和扩展,大规模图像集开始不断涌现。面对庞大的图像数据集,需要建立有效的图像描述、管理和检索机制。基于内容的图像检索顺应发展的需求,在图像资源管理方面的重要性日益突出。本文围绕基于内容的图像检索的关键技术展开讨论,重点研究了图像内容的描述方法、数据库索引技术、相似性匹配规则以及相关反馈机制。完成的工作主要包括:（1）构建图像特征数据库。以 caltech 标准图像库作为研究对象,结合时域和频域提取图像的颜色、纹理和形状特征作为图像内容的描述,构建特征数据库。提取的特征主要包括 HSI 中心矩法、直方图相交法、灰度-梯度共生矩阵纹理、和边缘投影直方图。（2）建立索引结构模型与特征匹配规则。初始检索设计了两种检索方式供用户选择,一种是通过建立数据库索引树结构 Kd-tree 进行近邻查询,另一种是线性检索,根据相似性度量准则计算示例图像的特征向量与特征数据库中特征向量的距离,按照相似程度输出与示例图像有相近特征的图像。（3）基于 IDEA 和 Sqlyog 软件平台,设计并实现了包含特征提取、数据库、索引匹配、查询显示和相关反馈五个功能模块的基于内容的图像检索实验系统。（4）利用标准图像库对实现的图像检索系统的功能进行测试。针对提取的不同特征进行实验,比较其检索效率,验证了有效的图像内容描述方法。
关键词: 图像检索; 基于内容; 图像索引
AbstractWith the continuous deepening and expansion of the application field of digital information, large-scale image collections began to emerge. Facing the huge image data set, it is necessary to establish an effective image description, management and retrieval mechanism. Content-based image retrieval conforms to the needs of development, and its importance in image resource management has become increasingly prominent. This paper discusses the key technologies of content-based image retrieval, focusing on the description methods of image content, database indexing technology, similarity matching rules and related feedback mechanisms. The completed work mainly includes:Construct an image feature database. Taking caltech standard image library as the research object, combining the time domain and frequency domain to extract the color, texture and shape features of the image as the description of the image content, and construct the feature database. The extracted features mainly include HSI central moment method, histogram intersection method, gray-gradient co-occurrence matrix texture, and edge projection histogram.Establish index structure model and feature matching rules. The initial retrieval is designed with two retrieval methods for users to choose. One is to perform nearest neighbor query by establishing a database index tree structure Kd-tree, and the other is linear retrieval, which calculates the feature vector of the sample image and the feature database according to the similarity measurement criterion. The distance of the feature vector, according to the degree of similarity, output images with similar features to the example image.Based on IDEA and Sqlyog software platform, a content-based image retrieval experiment system including five functional modules, including feature extraction, database, index matching, query display and related feedback, was designed and implemented.Use the standard image library to test the functions of the realized image retrieval system. Experiments on the different extracted features, comparing their retrieval efficiency, verify the effective image content description method.Keywords:image retrieval;content-based;image index