It's not that simple, you can't just look at the source pixel dimensions as the source file is likely to a hardware standard (e.g., 1920x1080). Web encoding can and generally should be cropped to the actual content, but mezzanine files generally are the standard dimensions of the production or editing workflow.

You scan for black bars, but you'll have to scan the whole movie, as sometimes content (usually documentaries) is presented in a mix of aspect ratios, parts with bars on top and bottom, parts with bars on sides, parts in full screen. So that's an expensive operation, taking you from one or two pass "off the shelf" encoding to three pass encoding where the first pass is this custom scan.

Note that the best encoding shops have a human make some judgment calls about the content, and set encoding parameters individually.