What is Query by Example?

Query by Example: Query by example is a friendly way, it is the most commonly used query method in CBIR. The disadvantage is that the user's understanding of the sample image may be different from that of the retrieval system. To avoid the impact of this situation on search results, sometimes additional attributes need to be added. ^[1]

This subsystem plays an important role in the CIBR system. Users submit query requests through this system. The final results of image retrieval are also returned to users for browsing through this subsystem. The common search methods provided by the user subsystem include the following:

1. Query by Example: The example query is a friendly way and is the most commonly used query method in CBIR. The disadvantage is that the user's understanding of the example image may be different from that of the retrieval system. In order to avoid the impact of this situation on search results as much as possible, sometimes additional attributes need to be added.

2. Query by Sketching: The system provides a window for users to draw sketches, and the system extracts features from the sketches and retrieves them. This method provides users with greater imagination, the disadvantage is that the compatibility is poor, and it will bring additional burden on the user.

3, Browse by Categories (Browsing by Categories): In some cases, users may not be familiar with the specific content of the image that they need to query, or the user is vague about the image that they are looking for. At this time, they can browse according to the system's classification system.

Example query status of foreign research and development

QBIC (Query By Image Content) is the most representative CBIR system developed by IBM. Provide users with a variety of search methods such as color, texture, sketch, and shape through a friendly interface, and provide a means to fuse multiple search features. Users can define the weight of each retrieval feature in the retrieval, so that they can control the retrieval direction according to their needs. The system conducts inquiries by way of example, and also provides browse and retrieval means. The system has the following two key characteristics: one is to use the color, texture, shape, motion, etc. of the images, videos and their objects as the key features of the query; the second is to have a graphical query language. Queries are presented through drawing, selection examples, and other graphical approaches.

When a user inputs an image, a sketch, or an image fragment, QBIC analyzes and extracts features such as color, texture, and motion changes of the input query image, and then performs different processing according to the query method selected by the user. The system also provides a variety of standard models, representing different colors, textures, and contour structures. The user can select the sample image that is closest to the object to be retrieved, perform a new search, and obtain a similar result image. The algorithm in this paper makes comprehensive use of color and texture features, and uses an example-based query method.

Example query status of domestic research and development

In recent years, great progress has been made in the research of image retrieval technology in China. For example, the image color-based retrieval system and image shape-based retrieval system studied by the Department of Computer Science of Zhejiang University, the prototype system of CBIR for still images on the Internet of Tsinghua University, and the Multimedia Information Retrieval System (Institute of Computer Technology, Chinese Academy of Sciences) , MIRES), etc. Among them, MIRES is a key laboratory of intelligent information processing of the Institute of Computing Technology of the Chinese Academy of Sciences, developed under the support of the National 863 Program; it can complete the retrieval of sample images, keywords, or two modes of interactive mixing in three ways, allowing users to set weights for various features Factors that support relevant feedback. ^[2]