A person who cannot see naturally cannot interpret graphical elements. A screen reader can read the corresponding “IMG” HTML code, but that usually hardly helps. This is especially the case when a web page is generated by a program, and the image files are all internally organized by “hashed” values. With a hand-crafted web site, perhaps the image file name (such as “mom-and-dad-christmas.jpg”) can still provide some hints. However, commercial web sites with thousands of bitmaps cannot use intuitive file names. For search efficiency, the filenames are often either sequential or “hashed”. In other case, it is likely to have nothing to do with the content of the bitmap.
Note that this also applies to the hyperlink (the href attribute) that an image may include. With a hand-crafted web site, again, it may not be too bad. The filename of an HTML document may suggest its content. However, with a dynamically generated web site, a link is usually a CGI (common gateway interface) script with a long list of parameters, such as http://www.amazon.com/Garmin-Pocket-Vehicle-Navigator-Integrated/dp/B000BKJZ9Q/sr=8-1/qid=1162837570/ref=pd_bbs_sr_1/104-5120226-5439905?ie=UTF8&s=electronics.