Image search engines

Today the most part of the web search engines including those that work with multimedia is based on tags and other text descriptions. It means that they search only for information that is described by text and marked and only for those users who search in the used markup language.
To create an image recognition based search is a very complicated mathematical task that no one can cope with yet. To be more exact, it is still complicated to design an image based search engine that will work better than a tag based search engine. And if it is not achieved the advantage of using such search engines will be unobvious.

That is why best results are shown by the web search engines that analyze all the information about an object: text description, graphics, speech and other sounds (if it is a video).

There are two main groups of image recognition based search engines:

1. search engines that do not recognize images but just use search technologies of objects that are similar by some characteristics.

2. search engines that analyze visual content

Technologies may combine: for example, the identity of images may be detected on the base of not only visual information but also text description.

In the first group one may reckon the following projects:

Like.com (технология Riya) — визуальный поиск для e-shopping. Применяется технология поиска похожих изображений Riya. Идентичность определяется по совпадению достаточно большого количества характеристик, присваиваемых изображению. Количество уникальных посетителей ежемесячно — более 2-ух миллионов человек.

Tineye.com is a beta version of the web search engine provided by Ideeinc, a company that develops advanced image identification and visual search software. The search is made on the base of some parts of image copy. The site has 27 288 unique users per month.

Picitup.com is a web search engine provided by an Israeli company. It works in the same way as Like.com; the search is made on the base of text description and visual information.

Tiltomo.com (beta version). The search is made in two ways: text description/colour/texture or 100%/colour/texture. A user can choose a search regime by himself. Thus, a content of the image is not analyzed.

Of course, search engines that can look at an object and say what exactly it shows are much more interesting. And then objects are compared on the base of the obtained information. It is the second group of search engines:

blinkx.com has 3 039 171 users per month. It recognizes images either by text description or visual information. It is reputed to be the best video search engine because it uses all the possible methods.

Betaface.com uses b2b base to search for images and videos. The service creates a base of individuals; its main virtue is that the service can recognize voice and speech. The company is going to release web API to distribute and design applications based on its runner.

Eyealike.com

uses several technologies: visual search, it analyzes and recognizes graphics, and on tags based text search. It also uses b2b. Three products are promoted:

1. Eyealike VisualAd. A platform to put up an advertisement that mainly uses for social nets; it takes into account an image content.
2. Eyealike Copyright. It is used by multimedia content producers to detect a proprietary content downloaded by users on such sites as Youtube and etc.
3. Eyealike Faces. It allows to search by images, recognizes faces and finds similar photos and people.

Delvenetworks.com offers a fully finished b2b decision to broadcast in the Internet and download videos. It also provides the service that search for similar multimedia objects by text description and audio or video content.