Version:           0.10 (using KDE 4.0.5)
Installed from:    Unspecified

in anticipation of digiKam 0.10 supporting XMP, i'd like to propose a new feature that will enable users to add more value to their images.
currently available metadata formats allow us to assign keywords to an image. i believe it's a good practice to name objects captured in a shot as keywords. however, in many cases just naming them is not enough. please take a look at this picture, for instance: http://farm4.static.flickr.com/3141/2749777162_60a5d02599_b_d.jpg. let's suppose this image has "Gilles Caulier" among its keywords. how are we supposed to find Gilles on the picture if we don't know him by sight? even if you're a good friend of him it'll take you quite a while to pinpoint Gilles' face in this crowd, and a positive result of the search is not guaranteed. so, why not make use of XMP and let users embed such data into new properties of existing XMP schemas or into a new XMP schema?
one of possible solutions is to add details of a square that can point to a certain area of an image to any object on the picture, like this:
http://www.geocities.com/rustahm/images/P5230296.JPG. these details can be embedded as XMP metadata as per this data diagram: http://www.geocities.com/rustahm/images/diagram.pdf, where ns is a new namespace [or an existing one], ns:object is an unordered array of stuctures, ns:position is a structure containing details of a square: coordinates of the left bottom corner [ns:x and ns:y] and length of its side [ns:sideLength].
this solution might be bad, but i hope it gave you the notion of my idea. once an image has this metadata, digiKam [or any other application] can use it to point to the area where the object is, if the user toggles the name of the object.