I'm getting increasingly hacked off with scientific publishing. Every journal I've ever submitted to has had serious issues with how it copes with figures.
These vary from the annoying (only accepting images in specific versions of proprietory formats) to the laughably amateurish (specifying either
the desired image sizes or the resolution - but never both).
For my most recent paper, I created a couple of complex and detailed scatter-plots. Mindful of issues such as colourblindness, I spent considerable time making these as clear as possible. These were then composited (with other panels) in high-res, at an appropriate size to show symbols representing different types of datapoint.
So during typesetting they rearrange the panels, shrinking the graphs down. But they don't save all that much space because for these the new layout wastes over a third of the available width. The other panels fare equally badly in the new layout; comparison between a matched set becomes much harder and they duplicate labelling to where it doesn't even make sense.
And after an acceptance process taking several months they give us 24 hours from Friday evening to review the proofs in a format we can't even see.
And then when it's done, they put the paper figures and a pdf on-line with the images converted to low-res jpegs.
So now, no-one can see a good version of the image.
Why can't it be different? There's no good reason for any of that.
I propose a journal which tries to honour the integrity of the figures with just as much care as is given to the paper's text. It tells authors what sizes of image it will accept, and at what resolutions. It gives advice about clarity both in 'guidance for authors' and throughout the acceptance process. It doesn't make, or require, changes at the final stage. It allows sufficient time for review of proofs. If the paper is available online (and it should be), figures are available at the resolution supplied.
 I'm not going to say what. But it's got a reasonable impact-factor.