[FFmpeg-devel] FFmpeg Small Task: Optimal Huffman Tables.

[Reimar Döffinger]

On 22 Apr 2012, at 20:36, briggers <briggers at ninthfloor.org> wrote:
> On Sun, 22 Apr 2012 19:29:25 +0200, Reimar Döffinger wrote:
>> On Sun, Apr 22, 2012 at 08:23:45PM +0530, briggers wrote:
>>> I expect to have quite a bit of free time over the next few months
>>> which I hope I can spend constructively by contributing to the
>>> FFmpeg project. I was looking through the small tasks list on the
>>> wiki at multimedia.cx [1] to get an idea where I could start. I was
>>> thinking I could start with the implementation of optimal huffman
>>> tables for (M)JPEG. I'll try to tread lightly, but I hope you'll
>>> forgive any future brain farts. Thanks.
>> 
>> Don't worry about mistakes, reviews are there for catching it
>> (but if you are new it helps if you can forgive if reviews seem/are
>> a bit harsh, people generally have only time to point out all that
>> is bad, even if most is good :-) ).
>> After reading the task I see I may have misunderstood it.
>> I guess you mean adding a optimal huffman table for each frame
>> individually (i.e. more the JPEG use case)?
>> However there are also MJPEG formats that have one huffman table
>> in the global extradata header, in which case a proper two-pass where
>> the whole file is encoded first and the optimal table is calculated
>> only in the second pass is necessary.
>> But even without that I am not sure this is exactly the most
>> straight-forward/easy of the tasks listed, but you are very
>> welcome to do it :-)
> 
> Thanks for the tips. If you have some other "small task" in mind that perhaps would be easier or more straight forward and would better serve as an introduction to contributing to FFmpeg, I'd happily work on that instead.

Well, it's also important that it is interesting enough for you to do.
Demuxers usually are easy, so for example the VIVO demuxer. There might be more.
I think there is also a another request in FFmpeg trac where I mostly reverse-engineered a rather simple format.
Unfortunately I forget the ticket number. If such a thing motivates you, it might make some user happy since there's no way to read them yet.
What also should be mostly simple is making the imagepipe demuxer autodetect the image format which would making it much more user-friendly to use (and make it possible for FFmpeg to open images without relying on the extension).
Do whatever you feel interesting, though particularly the last one I think has a good chance of giving first results very quickly, in case that motivates you.