Librarians care a lot about metadata, or data that describes the content of the materials at hand. It is this metadata that provide access points through which users can search and retrieve whatever it is that their hearts desire. This is true even for, say, ebooks whose content is entirely searchable because the content of a book may in fact be something not mentioned in its text at all. For example, Arthur Miller’s The Crucible is an allegory of the McCarthyism, but one would not find the terms “communism” or Joseph McCarthy in the book itself.
Metadata are indispensable for materials that librarians call “non self-describing”. The Crucible is one example of a work that not self-describing. Works that most clearly fall into this category, however, are visual and tactile materials: pictures, paintings, artifacts (think things that go into museums) . . . etc. Essentially, these kinds of do not “speak for themselves” (despite being worth a thousand words). This painting depict irises, but unless somebody has tagged it “irises”, there is no way to search for an image of that painting by using that keyword. Likewise, it is a painted by van Gogh, but unless somebody has created metadata describing the picture as such, one cannot search for “van Gogh painting” and hope to find it.
So the creation of metadata is an attempt to link visual materials with texts or words so that they can be retrieved using those words. Another approach altogether is to allow searches using not text but the images themselves.
There are two tools that I know of that allow for searches using visual cues. The first is the “Search by Image” function of Google Image Search. This requires that one possesses a file of an image to begin with, but once one has this “seed”, the tree of visual searches can grow. Try this out if you haven’t already.
Something that I didn’t try until quite recently (because I didn’t know it exists!) is QBIC. QBIC stands for Query by Image Content and it allows you to search using not an entire image (in case you don’t have one) but features from that image that you happen to remember. This means you can search for an image by specifying that it has, say, a lot of blue in it. Or an oval shape. Quite cool really. Do check it out!