Web search engines once limited input for search targets to text only. The new versions of search engine providers are supporting the emerging area of visual search, that is, searching for targets based on graphical data objects. This can include as input: pictures, videos, and 3D models. This type of graphical search uses Content Based Image Retrieval algorithms. The free public search engines currently support only static images as input for visual searches, so this article covers only that type of visual search.
Image searching is now of two varieties, standard browser searches and mobile searches. The first type requires the user to manually upload an image file, whereas a mobile search app can operate directly from pictures taken by a smartphone’s camera. With either platform, images are matched in two ways, graphics-only pattern matching, or searching for textual metadata associated with the images. The results found by visual searches can be of two types: traditional links or direct visual displays.
The actual algorithm for graphics matches is termed content-based image retrieval. The results come from complex comparisons of the components of a target image, such as shapes, colors, textures, and any other visual data available. This type of pattern matching is quite complex and computationally expensive, yet much more accurate and efficient than simpler metadata searches. There are visual search engines that combine both metadata and image graphics techniques, requiring the user to enter both keyword data and an image. A search employing both techniques searches for a number of metadata matches first, and then refines the search on the list of returned images with content-based matching to find the best possible matches to return to the user as ranked results.
Slyce.it is the leading visual search technology provider for retail businesses and manufacturer brands. Customers can find retail information about products or brands of which they have personally taken pictures. This is a dramatic shift in helping consumers quickly find a source for anything they see in the real world. Slyce’s visual product search technology is available to their retail partners to integrate into existing applications. They also have in-house consumer apps: SnipSnap, Pounce, and Craves to immediately supply ways of performing visual searches. Slyce is also applying their technology to other areas of use, to be announced in the future.
By integrating their visual search into retailers’ systems, the search results are customized to give product recommendations. A result screen may list identical items with buttons or links to directly purchase immediately, or offer prize contests for the user to be entered into as a contestant. One of the unique capabilities of Slyce’s retail software is that it can recognize products based on very little information, such as the type of materials and color of an item photographed.
Some visual search engines are Google Goggles by Google Labs, CamFind by Image Searcher, Inc., and LikeThat Apps Suite, including LikeThat Pets, LikeThat Dècor, and LikeThat Garden, by Superfish corporation. Clearly, visual search is the way of the future.