> But as the images are generated on the fly, the generated 
> image depends on the last visited webpage.

Yuk.. strange behaviour

> So my wish:
> There should be an option:
> 'Read linked images for the web page directly after 
reading 
> the web page'

Well, this might be interesting, but quite complicated to 
implement (mainly because the engine crawls on a heap 
basis, not a stack basis, and dig into the site layer per 
layer, deeper and deeper ; isolating specific areas would 
be quite complicated)