Introduction to the famous Image Retrieval System

[10:32:00 | by: If you have any feelings]

1. QBIC (query by image content) image retrieval system is the first content-based commercial image retrieval system developed by IBM in 1990s. The QBIC system provides a variety of query methods, including the use of standard charts (provided by the system itself) for retrieval, the user to draw a simplified graph or scan the input image for retrieval, and the selection of color or structure query methods, you can search for dynamic image clips and objects in the foreground. When a user inputs an image, a simplified image, or an image segment, QBIC analyzes and extracts the color, texture, shape, and other features of the input query image, then, different processing methods are performed based on the query method selected by the user. The color features used in QBIC include the color percentage and color location distribution. The texture features used are improved based on the texture representation proposed by Tamura, that is, it combines the characteristics of roughness, contrast and directionality. The shape features used include area, circular degree, eccentric degree, spindle bias and a set of algebraic moment constants. QBIC is one of the few systems that have considered high-dimensional feature indexes. In addition to content-based retrieval, QBIC also supports text query. For example, the San Francisco Museum of Modern Art provides a standard description of each piece: the author, title, date, and natural description of many works. 2. Virage is a content-based image retrieval engine developed by Virage. Like QBIC systems, Virage also supports Image Retrieval Based on visual features such as color, color layout, texture, and structure. Jerry and others further proposed an open framework for image management, which divides visual features into common features (such as color, texture, and shape) and domain-related features (such as for face recognition and cancer cell detection. Virage's visual information retrieval Image Engine provides four types of visual attribute retrieval (color, composition, texture, and shape ). Each attribute is assigned a weight of 0 to 10. Color retrieval is the simplest and clearer. The software analyzes the color, color, and fullness of the selected basic image, then, find the image closest to these color attributes in the Image Library. The composition property refers to the approximate degree of the relevant color areas. You can set one or more attribute weights to optimize the search. To achieve the optimal balance, repeat the experiment, but the retrieval process is quite fast. In the result display matrix, you can select 3, 6, 9, 12, 15, or 18. By adjusting the weights of the four attributes, different search results are displayed. A simplified chart is arranged in descending order of similarity. Click the title of the simplified image to get a detailed description of the image, including the similarity ratio calculated by Virage. 3. retrievalware is a content-based image retrieval tool developed by Excalibur Technology Co., Ltd. In earlier versions, we can see that the focus of the system is to use neural network algorithms for image retrieval. In the new version, R provides retrieval based on six image attributes: color, shape, texture, color structure, brightness structure, and aspect ratio. The color attribute is used to determine the color of the image and the proportion of the image, but does not include the determination of the color structure or position, which is controlled by the color structure attribute; shape attributes refer to the relative orientation, curvature, and contrast of the contour or line of an object in an image. Texture attributes refer to the smoothness or roughness of an image and the surface characteristics of an image; brightness attribute refers to the brightness of a pixel combination of an image. This is a powerful image retrieval tool. 4. PhotoBook is an interactive tool developed by the multimedia lab of the Massachusetts Institute of Technology for image query and browsing. It consists of three subsystems, which are responsible for extracting shape, texture, and facial features. Therefore, you can search Images Based on shapes, textures, and facial features in these three subsystems. In foureyes, the latest version of PhotoBook, Picard and others proposed the idea of adding users to the image annotation and retrieval process. At the same time, because people's perception is subjective, they propose a "model set" to combine human factors. Experimental results show that this method is very effective for interactive image annotations. 5. visualseek is a search tool based on visual features. webseek is a text or image search engine for WWW. Both retrieval systems are developed by Columbia University. They use spatial relationships between image regions and visual features extracted from the compressed domain. The system uses color sets and wavelet-based texture features. Visualseek supports both visual feature-based queries and spatial relationship-based queries. Webseek consists of three main modules: image/video acquisition module, topic classification and index module, and search, browse, and search module. Compared with other multimedia retrieval systems, visualseek has the following advantages: efficient Web image information retrieval, advanced feature extraction technology, powerful user interface, simple operation, and rich query methods, the output screen is vivid and allows users to download information directly. Webseek is an independent web visualization programming tool. It has catalogued 650000 images and 10000 image fragments. You can use directory browsing and feature retrieval to retrieve images.