I'd like to thank Eric Rescorla for making an excellent and pithy point about the purpose of publishing images of marked ballots. But first, thanks (again) to Mitch Trachtenberg of the Humboldt Transparency Project for publishing a hand-picked set of ballot images that provide a great example of the difficult borderline cases of interpreting hard-marked paper ballots -- whether it is a human or some software doing the interpreting. Ballot publication can show how much of a given election result actually depended on these borderline cases. Eric made a broader point that is so widely misunderstood that it truly merits repetition:
The main point of publishing ballot images is to allow people to independently verify that the images published correspond to the votes recorded for those images.
True, but verification requires more than ballot images - it requires that each image is published along with information about how that ballot's marks were interpreted as votes that were counted. I cringe every time somebody talks about ballot image publication as though it were just posting some JPEGs on a Web site.
By viewing the images plus these "cast ballot records", members of the public can look at a ballot image and decide for themselves whether they think its votes were interpreted correctly -- and if not, whether the putative mistakes are enough to effect the outcome of a race.
And just as important, consider the cases where an election official is involved in deciding an ambiguous mark - particularly at large scales such as with vote-by-mail. As a result, broader transparency requires that the election process maintain audit records of these decisions, in case they need to be re-visited.
So, sure, the publication of images alone is helpful for transparency, but Mitch's examples show how much interpretive leeway there can be. And in close elections, that leeway can influence whether a recount is required, or even influence an election result. So it's just as important to maintain and publish cast-ballot records, audit records, and the like.
But that is a lot of work!
And often is not feasible with current voting systems and election management technology. It's actually quite a job to maintain and publish all this information in a form useful to members of the public - a job that we're working on at the TrustTheVote Project of course, by building all of our election technology system components with the "Save Everything" principle.