Noindex with a canonical tag back to the unfiltered page? Why can't you do both?
Strikes me that noindex might in a way conflict with the canonical tag, particularly, if as describe, noindex causes the noindexed page (the one that would contain the canonical link) to be crawled less frequently over time.
I did some searching, and found that Google's John Mueller covers this question here, as Google's crawling process enters into it, and there is indeed a kind of conflict if you use both... Canonical conflicts with noindex? http://productforums.google.com/d/msg/webmasters/0sqRrolO_Ss/igOdQIjwKdEJ
... Before the rel=canonical link element was announced, using noindex robots meta tags was one way that webmasters were directing us towards canonicals, so this is certainly something we know and understand. However, with the coming of the rel=canonical link element, the optimal way of specifying a canonical is (apart from using a 301 redirect to the preferred URL) is to only use the rel=canonical link element.
One reason for this is that we sometimes find a non-canonical URL first. If this URL has a noindex robots meta tag, we might decide not to index anything until we crawl and index the canonical URL. Without the noindex robots meta tag (with the rel=canonical link element) we can start by indexing that URL and show it to users in search results. As soon as we crawl the canonical URL, we can change to the canonical URL instead. It's also much safer because you don't have to worry about serving different versions of the content depending on the exact URL :-).
There's a bit more in John's two posts in the discussion worth looking at.
There's also a Matt Cutts video on this page... About rel="canonical" https://support.google.com/webmasters/answer/139394?hl=en
At about 7:04 into the video, Matt talks about about different sort orders in ecommerce, but not much about filtering (I haven't watched the whole video in quite some time, so he might).
Much more to be said about keeping it simple... and also making sure that you're not filtering out significant amounts of information with your filters. Make sure that you are distinguishing between attributes (things like colors) and categories (things like brands that you want indexed on their own). So, the solutions we're discussing, IMO, probably wouldn't work for filtering brands... but I'd love to hear otherwise if they will.
That said, there is a video on the canonical and pagination by Maile Ohye, not quite the same thing as what's being discussed here, that I've linked to several times and you can find by site search... and it occurs to me that conceivably a paginated approach, with the canonical, might work as a way to both sort and then filter by brands... if someone wants to work that one out ;). (I haven't thought it through, but it could be a remote possibility.)
Also, there are other ways of filtering faceted searches, but all probably more complex than aakk9999's elegant suggestion of the dummy folder.