The <art> element enables associating metadata and a descriptive title with references to images or other media objects. It's primary purpose is to serve as a "managed object" in a re-use or content management context.

Because the base <image> element doesn't allow <data> elements in its content there's no direct way to associate classifying or descriptive metadata with <image> elements. Thus the <art> element provides a way to do it.

The <art> element is for use in block contexts (anywhere <p> is allowed). The <art-ph> element is for use in phrase contexts (anywhere <ph> is allowed).

With DITA 1.3 the <art> element could be specialized from <div>, which would remove the need for separate <art> and <art-ph> element types.