Content Based Image Retrieval

Under Content Based Image Retrieval ( CBIR ) means a content-based image search. Alternative names are query by image content ( QBIC ) and content-based visual information retrieval ( CBVIR ). This is a special field of image processing and the retrieval of information ( information retrieval ) in large databases (see also for a current overview).

"Content-based" means an analysis of the current content of an image, i.e. the colors, outlines, surfaces (textures) or other information (so-called feature vectors ) that can be determined using automatic image processing. (see also for an introduction) The task of the image search is to sort a list of existing images in such a way that the images you are looking for (e.g. using a reference image) are as close to the front as possible. Images are sorted based on their similarity to the reference image, which is determined by a distance function and the feature vectors of the images. A quality measure assesses the sorting , which depends largely on the choice of the feature vectors and the degree of similarity.

In contrast to the content-based search, there is the “keyword-based” or “text-based” image search, in which the image is described (e.g. which persons are visible, which object is depicted or the coordinates of the images) (see Information Retrieval ).

The content-based image search is used for image databases, in the field of medical image processing and when searching for plagiarism (near duplicate detection).

In the area of ​​web search engines, the term "reverse image search " or a "visual search engine" is also used, depending on the provider .

See also

• Image recognition
• Edge detection
• Object recognition
• Image search engine
• TinEye

Web links

• Overview of concrete CBIR implementations: Content-Based Image Retrieval Systems: A Survey (PDF; 2.3 MB), Remco C. Veltkamp, ​​Mirela Tanase, Utrecht University, Technical Report UU-CS-2000-34, 2000 (revised and extended version, October 2002)
• Open source CBIR tool: The GNU Image-Finding Tool
• Open source CBIR library for Java: Lucene Based Image Retrieval - LIRE

Individual evidence

1. ↑ Content-based Multimedia Information Retrieval: State of the Art and Challenges (PDF; 169 kB), Michael Lew et al., ACM Transactions on Multimedia Computing, Communications and Applications, Vol. 2, No. 1, February 2006, pp. 1-19.
2. ^ Image Retrieval: Ideas, Influences, and Trends of the New Age , Ritendra Datt, Dhiraj Joshi, Jia Li, James Z. Wang, ACM Computing Surveys, Vol 40 (2), Article 5, April 2008.
3. ↑ Visual Information Retrieval using Java and LIRE , Mathias Lux, Oge Marques, Synthesis Lectures on Information Concepts, Retrieval, and Services 2013 5: 1, 1–112
4. ↑ Jing Huang, S. Ravi Kumar, Mandar Mitra: Combining supervised learning with color correlograms for content-based image retrieval . In: MULTIMEDIA '97 Proceedings of the fifth ACM international conference on Multimedia . ACM , New York, NY, USA 1997 ( ACM , PDF ).
5. ↑ A review of content-based image retrieval systems in medical applications — clinical benefits and Future Directions , Henning Müller et al., International Journal of Medical Informatics, Vol 73 (1), pp. 1-23, 2004.
6. ↑ Joshua Lockhart: Top 6 Visual Search Engines for finding the image you want . makeuseof.com. 1st of May 2013.