Lets say a page on example.com has all of its images coming from img.example.com:
<img src="http://img.example.com/keyword.jpg" alt="keyword"/>
while robots.txt on img.example.com disallows crawling:
User-agent: *
Disallow: /
Would this make such images contribute less to a keyword value comparing to images from allowed domains?
I would guess so, because google probably would want to crawl an image url to see if anyone is using it on other sites, or if it even exists.