DDD System and DDD Design, two Hamburg-based software companies, presented MediaLens Photo, a new way of searching and finding images, at the third PICTA Picture Agency Trade Fair in Hamburg, April 1-3.
This is absolutely a totally different and interesting approach regarding the design and functionality of User Interfaces. You have to enter the website to get a first visual impression of the interface - it is really difficult to describe it only in words.
According to DDD and the CEO Matthias Godenrath MediaLens Photo is using Fuzzy-Logic-Searchtechnology for the retrieval of images: "Finding with [image] attributes and properties instead of searching with keywords."
Using sliders (yes!) for image attributes like age, haircolor, cultural origin, temper, gregariousness, humour, sensuality etc. Art Buyer and Picture Editors can limit the degree of expression of each of these image attributes according to their imagination how the desired but still unknown picture should look like. This search by sliders is also available for the technical details of the image (size, orientation etc.). Using the "Advanced Search" the desired color and brightness can be set separately for the fore- and background. Using the "Focus"-feature one can determine with a mouse-click where the music should play, i.e. where within the desired but still unknown image a person or the action should be or take place - to the left, the right, at a corner etc; you can even determine the area of the focus - big, small etc.
Others search technologies are often working with an intial keyword search and/or reference images. DDD explains that a normal search with keywords often reveals unsatisfying search results, especially in the case of ambigous keyword phrases. Sometimes the researcher even does not have a clear imagination of what he is really looking for, only an idea in the distance being very vaguely. They claim that thanks to their Fuzzy-Logic-Searchtechnology the search results presented by MediaLens Photo are towered above and superb because their technology is able to deal with the entered "fuzzy-data" being ambigous and sometimes even contradictory. MediaLens Photo also offers the possibility to search for "similar" pictures called the "more like this"-feature.
The technology shown at the PICTA was a first stand-alone-prototype, not Internet-ready and only with a few thousand images. Interested users can now directly surf to ml.ddd.de and request a password to "play" with the technology.
To our knowledge DDD System/DDD Design are newcomers in the Stock Photo Industry. However they have another sophisticated product called MediaLens Music allowing users to search for music titles with the same Fuzzy-Logic-Searchtechnology. From my point of view at the present development stage MediaLens Photo will be a benefit mainly to researchers of portrait photography in the Royalty Free Sector. The User Interface can be customized and re-adjusted to the needs of any customer agency, even extended with more functionalty. According to CEO Matthias Godenrath the idea for MediaLens Photo was born in a photo agency in New York.
Quite interesting would be the scientific background behind this technology and if its working with Content Based Image Retrieval (CBIR; colour, shape, texture etc.) or Keyword-Related Image Retrieval (KRIR) or a combination of both methods (to top the abbreviation mania, let´s call the latter COMB). As I have sympathies for both techniques, keep in mind, my "job" in a weblog, as I see it, is not to protect either side of the the table.
We have seen a lot of companies -each naturally offering the next big thing- dealing with Image Retrieval in both the CBIR- and the KRIR-way in the past four years: some changed their business model to brand monitoring, law enforcement and content security (Cobion, recently sold), some pretty active (Ask Angela and the "magically related" images by Comstock; LTU), some being the real masters of the universe (Idee Inc.), others gone (Piaxa) and still! new projects (Minitera.ru, Project ImEx). And of course Imagepoint.biz. [This is not a full listing of all companies active in this field: in addition to diddly.com and ditto.com - mentioned below - there are a bunch of others active in this field like VIMA Technologies or Pixlogic]
These companies and their search technologies can mainly be divided in four categories: (a) searching an intranet database or the internet, (b) working with CBIR or KRIR or COMB, (c) giving the possibility to "learn" on query results by interactive query refinement and to show "similar" pictures and (d) used in the Stock Photo Industry, i.e. business dealing with images, or in consumer photo databases like online-print-services with their huge amount of pictures.
Keep in my mind that all these technologies are not about selling significantly more pictures or expanding the market size of the Stock Photo Industry. They are all about selling the more appropriate picture. Independently from their scientific background and approach only a few of these companies introduced fine-tuned and state-of-the-art developed User-Interfaces. This is where MediaLens Photo admittedly may demonstrate a huge progress for Art-Buyers and Picture Editors.
