It's time to start work on DGDeblock!
First, I would like to ask for your torture clips and your recommended filters that I can look at for inspiration and for comparative testing. I have some ideas about how to implement something that will be fast and effective but before I share it, let's hear from you.
DGDeblock
Re: DGDeblock
Thanks !
Footy season starts Sunday, seems very likely I'll have one then
Footy season starts Sunday, seems very likely I'll have one then
I really do like it here.
Re: DGDeblock
At least for the first few games this year (not my team yet), the broadcasters have been using better bitrates and it's less blocky although still otherwise mangled. I bet they lower the bitrate after Round 1.
Anyway, there are 3 576i clips in https://drive.google.com/open?id=0B5RV2 ... m1tYnJhejg
At circa the 8.1 second point in "SAMPLE3-WAFL-quite-blocky.mpg" it gets a tad blocky.
(Yes, in sample2 the guy's nickname is "basher".)
I can probably find some more if you need.
Anyway, there are 3 576i clips in https://drive.google.com/open?id=0B5RV2 ... m1tYnJhejg
At circa the 8.1 second point in "SAMPLE3-WAFL-quite-blocky.mpg" it gets a tad blocky.
(Yes, in sample2 the guy's nickname is "basher".)
I can probably find some more if you need.
I really do like it here.
Re: DGDeblock
Thank you, hydra3333. I'm going to implement DGFieldDeinterlace first.
Re: DGDeblock
DGFieldDeinterlace? Sounds juicy. Just curious, what kind of quality are you aiming for? I wish for something that can match QTGMC but with more speed of course. Would be amazing if you can pull it offadmin wrote:Thank you, hydra3333. I'm going to implement DGFieldDeinterlace first.
Re: DGDeblock
Greetings, astrayred!
At first I will just make a fast CUDA version of FieldDeinterlace(), which is an adaptive deinterlacer that works pretty well but is not motion-compensated, etc. Meanwhile I am working on CUDA motion estimation/compensation. If that pans out, then a motion-compensated deinterlacer would be feasible. We will see. Meanwhile, I have good news about DGDeinterlace. I will post about it shortly.
At first I will just make a fast CUDA version of FieldDeinterlace(), which is an adaptive deinterlacer that works pretty well but is not motion-compensated, etc. Meanwhile I am working on CUDA motion estimation/compensation. If that pans out, then a motion-compensated deinterlacer would be feasible. We will see. Meanwhile, I have good news about DGDeinterlace. I will post about it shortly.
Re: DGDeblock
Awesome sir! Can't wait to try it out!