Print

Print


Hi, Chris!

Thanks for the response, I really appreciate it. We were just about to experiment with changing that "L"  to "RGB" on our test server.  What you're suggesting sounds fantastic--I was concerned about the potential performance impact of sending everything (even the stuff for which "L" is accurate) through under RGB mode. We have a mix of grayscale/color content in the same instance of Chronam.

We'd definitely be willing to help test this, if you think it's a change others could benefit from--we'd apply your changes as a patch.

Thanks again--

Stephanie

--
Stephanie Williams
Digital Projects Programmer
[log in to unmask]<mailto:[log in to unmask]> | 919-962-4836
North Carolina Digital Heritage Center | digitalnc.org<https://outlook.unc.edu/owa/redir.aspx?C=N-8nuKR7jUS9PaaVyqNqbXTv1DlUr9AIj55oMDWUJwwfNlPnRpIW5sSsu0vdD50GZ1KfqnWxtDE.&URL=http%3a%2f%2fdigitalnc.org%2f> | @ncdhc<https://outlook.unc.edu/owa/redir.aspx?C=N-8nuKR7jUS9PaaVyqNqbXTv1DlUr9AIj55oMDWUJwwfNlPnRpIW5sSsu0vdD50GZ1KfqnWxtDE.&URL=https%3a%2f%2ftwitter.com%2fncdhc>
________________________________
From: Data, API, website, and code of the Chronicling America website [[log in to unmask]] on behalf of Chris Adams [[log in to unmask]]
Sent: Wednesday, June 24, 2015 2:10 PM
To: [log in to unmask]
Subject: Re: Color Images in Chronam?

On Jun 23, 2015, at 12:14 PM, Williams, Stephanie <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Thanks to Mike, I knew this was likely something to do with our JP2 library; I'm now pretty sure I've traced it back to a hard-coded frombuffer() setting--"L", for grayscale--in the NativeImaging module that Chronam uses to optimize image generation/broker image libraries.  I'm wary of the consequences of changing that, and will be investigating further.

Ah, this spot:

https://github.com/acdha/NativeImaging/blob/bad7425766bf1a4fd28452caf0afde9ab28d7a9d/NativeImaging/backends/aware.py#L220
https://github.com/acdha/NativeImaging/blob/bad7425766bf1a4fd28452caf0afde9ab28d7a9d/NativeImaging/backends/aware_cext.py#L105

That code was developed to support Chronicling America which at the time didn�t have color content. It shouldn�t be too invasive to do something like use the number of channels and bits-per-pixel values to select the mode from the list at http://pillow.readthedocs.org/en/latest/handbook/concepts.html#concept-modes. Unfortunately I don�t easily have access to Aware to test this but if someone else does I can merge the change and push it out to PyPI quickly.

Chris