Merging into one regex expression is hard here because POSIX regex does not support lazy quantifiers.
With GNU sed, you can pass the command as
sed 's/.*FROM \(.*\) as.*/\1/;s/FROM //' file
See this online demo.
However, if you have a GNU grep you can use a bit more precise expression:
#!/bin/bash
s='FROM some_registry as registry1
From another_registry'
grep -oP '(?i)\bFROM\s+\K.*?(?=\s+as\b|$)' <<< "$s"
See the online demo. Details:
(?i) - case insensitive matching ON
\b - a word boundary
FROM - a word
\s+ - one or more whitespaces
\K - "forget" all text matched so far
.*? - any zero or more chars other than line break chars as few as possible
(?=\s+as\b|$) - a positive lookahead that matches a location immediately followed with one or more whitespaces and then a whole word as, or end of string.