add dts extractor #61 #81
Conversation
|
First, thank you! I don't own any cameras that produce B frames, so...
and...
I don't really either and need to make another pass through to develop that understanding and compare this technique with gstreamer's |
src/codec/h264/dts_extractor.rs
Outdated
| r.read_u32((log2_max_frame_num_minus4 + 4).into(), "frame_num") | ||
| .map_err(DtsExtractorError::BitReader)?; | ||
|
|
||
| if let FrameMbsFlags::Fields { |
There was a problem hiding this comment.
- More specific comment: If
let Enum::VariantA { .. } = Enum::VariantBis dead code. Looks like the right side here should come from a field in the sps; maybe plumb it through as a parameter? - Broader comment: can we use
h264_reader::nal::slice::SliceHeaderinstead of this function's parsing logic? might keep things more readable.
There was a problem hiding this comment.
SliceHeader::from_bits() takes SPS and PPS as arguments, I'm not sure parsing PPS in parse_sps_and_pps and plumbing it in would be more readable.
|
The gstreamer implementation seems to add latency. I'm not sure if it's just some initial delay like with the MediaMTX implementation but otherwise that's a problem.
|
src/codec/h264/dts_extractor.rs
Outdated
| // Allows to extract DTS from PTS. | ||
| #[derive(Debug)] | ||
| pub struct DtsExtractor { | ||
| prev_dts_filled: bool, |
There was a problem hiding this comment.
More idiomatic to use prev_dts: Option<i64> rather than the separate field.
There was a problem hiding this comment.
Huh, it seems like extract_inner() relies on the prev_dts zero value.
|
Found an explanation for how this algorithm works: |
|
Would it be better to first add the dts extractor as separate module that people can use manually before integrating it into the |
This is a rust port of the gortsplib dts extractor. I don't really understand how it works, but my test stream doesn't stutter anymore.
I know you're busy, I don't expect this to be reviewed anytime soon.