Chapter 18

A Java applet version of the C-BIRD system can be found at

http://www.cs.sfu.ca/cbird/

A report on the state-of-the-art in Multimedia Search Engines is at

http://chorusgapanalysis.wetpaint.com/

A demo of QBIC as an artwork server is

http://www.hermitagemuseum.org/fcgi-bin/db2www/qbicSearch.mac/qbic?selLang=English

A demo version of the Alexandria Digital Library is available from

http://www.alexandria.ucsb.edu/adljigi/tutorials/walkthrough1/

A demo of the Berkeley Digital Library Project is at

http://elib.cs.berkeley.edu/photos/

A demo version of VisualSEEk can be reached at

http://www.ctr.columbia.edu/VisualSEEk/

The Informedia project has provided the search engine power for a commercially available system, at MediaSite, now rebranded as Sonic Foundary Media Systems

http://www.sonicfoundry.com/systems/mslive.asp

A demo of the NETRA system is available at web URL

http://maya.ece.ucsb.edu/Netra/netra2.html

The idea is to select an image, then a particular segment within an image, and search on that model.

A demo of Photobook is available at

http://www-white.media.mit.edu/~tpminka/photobook/

A free download version of Photobook is available at

ftp://whitechapel.media.mit.edu/pub/photobook/

A demo of the Visual RetrievalWare system:

http://vrw.convera.com:8015/cst

A video describing the technology for the Virage system is at web URL

http://www.virage.com/products/

under ``View Solutions Video". Virage provides the search engine for AltaVista's Image Search.

VIPER has a demo web site at

http://viper.unige.ch/demo/

The demo written in PHP is the lightest-weight version there (Java and CGI versions, as well, are available). The demo asks for a reference URL for an image (e.g., http://www.cs.sfu.ca/people/images/mark.gif). For that example image it proceeds to recover as many as 50 fairly irrelevant images, with the first potentially useful image appearing as number 49. Re-launching the query by marking the search result images as either relevant, non-relevant, or neutral, we are then presented with only 50 semi-irrelevant images. Honing the query again then leads to several possible useful images.

An excellent list of CBIR URLs is at

http://www.aa-lab.cs.uu.nl/cbirsurvey/cbir-survey/

A very useful description of issues amd challenges in CBIR is at

http://www.theopavlidis.com/technology/CBIR/index.htm

Standard Image and Video collections:

The National Institute of Standards and Technology has created what it proposes as a standard set of test videos for testing video segmentation and retrieval programs. The website

http://programs.researchchannel.com/displayseries.asp?collid=115#574

allows one to see these videos, and they can be purchased at

http://www.nist.gov/srd/nistsd26.htm.

A standard image set is the Corel Gallery of images, grouped by categories.

http://www.corel.com.

Categories are: accessories, active lifestyles, Africa, aliens, alphabet, amusement, anatomy, angels and spirits, animals, artists, arts and entertainment, babies, best wishes, bicycles, bird illustrations, boats, bowling, business, butterflies, camping, cars, cartoons, cats, celebrations, Celtic, child dreams, Chinese mythology, Christmas, circus, city life, cityscapes, clowns, dance, dark ages, dinosaurs, dogs and cats, education, Egypt, entertainment, everyday life, fantasy, farm, fashion, finance, fish, folk dances, food, forests and trees, fruits, fun, games, gardens, gothic, grandparents, great white north, hair, Hawaii, high tech, Hispanic, historical, history, hobbies, hockey, holidays, home, horses, infants, insects, interiors, international business, international cuisine, Internet, Inuit, Japanese culture, kids, kitchen, knights of the round table, landscapes, magic, martial arts, medicine, metaphors, millennium, mothers and daughters, music, new baby, occupations, office, opera divas, oriental rugs, outdoors, Paris life, patterns, people, pets, prehistoric, quilt letters, real estate, recreation, religion, satellites, school, science fiction, sea cartoons, sea creatures, seasons, small business, snow, space, special occasions, sports, sports figures, stained glass, stamps, still life, summer, textures, theater, tools, toys, travel, treats, trucks, vacation, weather, woodcut, workplace, and world people.

The keyframe generation method discussed in this chapter, and the videos used, are given at the website

http://www.cs.sfu.ca/~mark/ftp/AcmMM00/.

Further progress in image retrieval has generally been in the direction of incorparating semantic information into queries:

• http://www.svcl.ucsd.edu/projects/imgnote/
makes use of "semantic retrieval" -- the use of a training database of images with each image annotated with a natural-language caption. A system can then learn to form a map between words and visual parts of an image.
A problem with using natural-language semantics is dealing with the limited vocabulary in the training set. Instead, Query by Semantic Example learns a probabilistic model for each concept and represents each image by a distribution over the space of semantic concepts, http://www.infolution.com/products/roadahead.htm thus being able to make inferences using semantic spaces:
• http://www.svcl.ucsd.edu/projects/qbse/index_removed.htm