The problem is the nature of the medium. Not the maturity of the industry.
In photography, search is most often not for a particular object or product, but for a concept, which may be expressed by the state of a particular object.
(Continued after the Jump)
For example, If you want to find a dog photo, not a particular photo but a photo that expresses a certain concept, there is no possible answer (other than wading) but reliance on keywords. And when you do that search, if the results have not been edited for quality, you are going to waste time wading through though thousands of images to find a gem. Supplementary keywords added by viewers might improve keywording accuracy, however, it will be decades - at least - before a computer can parse the difference between a person smiling and showing their teeth - and that same person with a angry teeth-barring scowl, not to mention the micro-facial-expressions that differentiate happiness as compared to attraction.
Visual search can help with color, orientation, pattern, but not concept.
Image recognition can help with finding an exact image if you already have one, or an image with similarly structured feature points if you already have the image that you are looking for.
Object recognition can help with finding photographs of clearly delineated, unambiguously rendered objects, such as a bicycle, but will have a tough time distinguishing between a terrier and a cat. This will improve with time, but it takes a pair of eyes and a brain to distinguish subtleties, and this will not change. A combination of object recognition and contextual search will yield better results than either type of search alone.
But if you are conducting a search at a search engine like google, the results are images IN USE, not images FOR SALE. Big difference.
The problem is, there is almost always, as those attempting to solve licensing problems, a failure to recognize the fundamental differences between different types of products, and different types of intellectual property.
Searching for music is easier than searching for photos because there are relatively *few songs*. Searching for music -- a "mature" industry -- is a big headache, if you don't know the artist or the name of the song, but are just looking for a song that expresses a certain concept. However, lyrics can be of some assistance, beats per minute, genre, and so on. Yet, in the end, to be able to actually experience - by listening - to a song, will provide you with the final answer. So too, by seeing concepts in images, you really do have to either see it, or have had someone see it and keyword in conceptual keywords.
Heller, among others, suggest that even the big fish (Getty/Corbis/etc) can't/won't even attempt to solve these problems, because their results are absent images they don't represent. Yet, as noted above, it's more than that - these results would not include images that are for sale, yet not in a stock agency or indexed by Google.
Further, the licensing of images is a challenge too. If I told you that someone needed an aerial photograph of a football stadium, but they only wanted 200 copies for a printed brochure, how would you price that? Well, it's one thing if the brochure is for a local realtor to show how protected a home's traffic access is to the seasonal traffic jams on nearby roads, and completely another if that brochure will be what sells the NFL owners on having the next Super Bowl at that stadium.
Photography (and illustrations, for that matter) will continue to be far too complex and fractured an industry for there to be any one place to centralize the licensing of it
Please post your comments by clicking the link below. If you've got questions, please pose them in our Photo Business Forum Flickr Group Discussion Threads.