Anyway to clarify this code? (dictionaries)

[javuchi]

I've been searching thru the library documentation, and this is the
best code I can produce for this alogorithm:

I'd like to return a dictionary which is a copy of 'another' dictionary
whoes values are bigger than 'x' and has the keys 'keys':

def my_search (another, keys, x):
        temp = another.fromkeys(keys)
        return dict([[k,v] for k in temp.keys() for v in temp.values()
if v>=x])

Is there any way to improve this code?
I want to avoid converting the dictionary to a list and then to a
dictionay. Are there speed penalties for such a conversion?

Bye.