moonfire-nvr/server/src/h264.rs
2021-06-09 14:36:14 -07:00

439 lines
18 KiB
Rust

// This file is part of Moonfire NVR, a security camera network video recorder.
// Copyright (C) 2021 The Moonfire NVR Authors; see AUTHORS and LICENSE.txt.
// SPDX-License-Identifier: GPL-v3.0-or-later WITH GPL-3.0-linking-exception.
//! H.264 decoding
//!
//! For the most part, Moonfire NVR does not try to understand the video codec. However, H.264 has
//! two byte stream encodings: ISO/IEC 14496-10 Annex B, and ISO/IEC 14496-15 AVC access units.
//! When streaming from RTSP, ffmpeg supplies the former. We need the latter to stick into `.mp4`
//! files. This file manages the conversion, both for the ffmpeg "extra data" (which should become
//! the ISO/IEC 14496-15 section 5.2.4.1 `AVCDecoderConfigurationRecord`) and the actual samples.
//!
//! See the [wiki page on standards and
//! specifications](https://github.com/scottlamb/moonfire-nvr/wiki/Standards-and-specifications)
//! for help finding a copy of the relevant standards. This code won't make much sense without them!
//!
//! ffmpeg of course has logic to do the same thing, but unfortunately it is not exposed except
//! through ffmpeg's own generated `.mp4` file. Extracting just this part of their `.mp4` files
//! would be more trouble than it's worth.
use byteorder::{BigEndian, ByteOrder, WriteBytesExt};
use failure::{bail, format_err, Error};
use std::convert::TryFrom;
// See ISO/IEC 14496-10 table 7-1 - NAL unit type codes, syntax element categories, and NAL unit
// type classes.
const NAL_UNIT_SEQ_PARAMETER_SET: u8 = 7;
const NAL_UNIT_PIC_PARAMETER_SET: u8 = 8;
const NAL_UNIT_TYPE_MASK: u8 = 0x1F; // bottom 5 bits of first byte of unit.
// For certain common sub stream anamorphic resolutions, add a pixel aspect ratio box.
// Assume the camera is 16x9. These are just the standard wide mode; default_pixel_aspect_ratio
// tries the transpose also.
const PIXEL_ASPECT_RATIOS: [((u16, u16), (u16, u16)); 6] = [
((320, 240), (4, 3)),
((352, 240), (40, 33)),
((640, 352), (44, 45)),
((640, 480), (4, 3)),
((704, 480), (40, 33)),
((720, 480), (32, 27)),
];
/// Get the pixel aspect ratio to use if none is specified.
///
/// The Dahua IPC-HDW5231R-Z sets the aspect ratio in the H.264 SPS (correctly) for both square and
/// non-square pixels. The Hikvision DS-2CD2032-I doesn't set it, even though the sub stream's
/// pixels aren't square. So define a default based on the pixel dimensions to use if the camera
/// doesn't tell us what to do.
///
/// Note that at least in the case of .mp4 muxing, we don't need to fix up the underlying SPS.
/// SPS; PixelAspectRatioBox's definition says that it overrides the H.264-level declaration.
fn default_pixel_aspect_ratio(width: u16, height: u16) -> (u16, u16) {
if width >= height {
PIXEL_ASPECT_RATIOS
.iter()
.find(|r| r.0 == (width, height))
.map(|r| r.1)
.unwrap_or((1, 1))
} else {
PIXEL_ASPECT_RATIOS
.iter()
.find(|r| r.0 == (height, width))
.map(|r| (r.1 .1, r.1 .0))
.unwrap_or((1, 1))
}
}
/// Decodes a H.264 Annex B byte stream into NAL units. Calls `f` for each NAL unit in the byte
/// stream. Aborts if `f` returns error.
///
/// Note `f` is called with the encoded NAL form, not the RBSP. The NAL header byte and any
/// emulation prevention bytes will be present.
///
/// See ISO/IEC 14496-10 section B.2: Byte stream NAL unit decoding process.
/// This is a relatively simple, unoptimized implementation.
///
/// TODO: detect invalid byte streams. For example, several 0x00s not followed by a 0x01, a stream
/// stream not starting with 0x00 0x00 0x00 0x01, or an empty NAL unit.
fn decode_h264_annex_b<'a, F>(mut data: &'a [u8], mut f: F) -> Result<(), Error>
where
F: FnMut(&'a [u8]) -> Result<(), Error>,
{
let start_code = &b"\x00\x00\x01"[..];
use nom::FindSubstring;
'outer: while let Some(pos) = data.find_substring(start_code) {
let mut unit = &data[0..pos];
data = &data[pos + start_code.len()..];
// Have zero or more bytes that end in a start code. Strip out any trailing 0x00s and
// process the unit if there's anything left.
loop {
match unit.last() {
None => continue 'outer,
Some(b) if *b == 0 => {
unit = &unit[..unit.len() - 1];
}
Some(_) => break,
}
}
f(unit)?;
}
// No remaining start codes; likely a unit left.
if !data.is_empty() {
f(data)?;
}
Ok(())
}
/// Parses Annex B extra data, returning a tuple holding the `sps` and `pps` substrings.
fn parse_annex_b_extra_data(data: &[u8]) -> Result<(&[u8], &[u8]), Error> {
let mut sps = None;
let mut pps = None;
decode_h264_annex_b(data, |unit| {
let nal_type = (unit[0] as u8) & NAL_UNIT_TYPE_MASK;
match nal_type {
NAL_UNIT_SEQ_PARAMETER_SET => sps = Some(unit),
NAL_UNIT_PIC_PARAMETER_SET => pps = Some(unit),
_ => bail!("Expected SPS and PPS; got type {}", nal_type),
};
Ok(())
})?;
match (sps, pps) {
(Some(s), Some(p)) => Ok((s, p)),
_ => bail!("SPS and PPS must be specified"),
}
}
/// Parsed representation of ffmpeg's "extradata".
#[derive(Debug, PartialEq, Eq)]
pub struct ExtraData {
pub entry: db::VideoSampleEntryToInsert,
/// True iff sample data should be transformed from Annex B format to AVC format via a call to
/// `transform_sample_data`. (The assumption is that if the extra data was in Annex B format,
/// the sample data is also.)
pub need_transform: bool,
}
impl ExtraData {
/// Parses "extradata" from ffmpeg. This data may be in either Annex B format or AVC format.
pub fn parse(extradata: &[u8], width: u16, height: u16) -> Result<ExtraData, Error> {
let raw_sps_and_pps;
let need_transform;
let ctx;
let sps_owner;
let sps; // reference to either within ctx or to sps_owner.
if extradata.starts_with(b"\x00\x00\x00\x01") || extradata.starts_with(b"\x00\x00\x01") {
// ffmpeg supplied "extradata" in Annex B format.
let (s, p) = parse_annex_b_extra_data(extradata)?;
let rbsp = h264_reader::rbsp::decode_nal(&s[1..]);
sps_owner = h264_reader::nal::sps::SeqParameterSet::from_bytes(&rbsp)
.map_err(|e| format_err!("Bad SPS: {:?}", e))?;
sps = &sps_owner;
raw_sps_and_pps = Some((s, p));
need_transform = true;
} else {
// Assume "extradata" holds an AVCDecoderConfiguration.
need_transform = false;
raw_sps_and_pps = None;
let avcc = h264_reader::avcc::AvcDecoderConfigurationRecord::try_from(extradata)
.map_err(|e| format_err!("Bad AvcDecoderConfigurationRecord: {:?}", e))?;
if avcc.num_of_sequence_parameter_sets() != 1 {
bail!("Multiple SPSs!");
}
ctx = avcc
.create_context(())
.map_err(|e| format_err!("Can't load SPS+PPS: {:?}", e))?;
sps = ctx
.sps_by_id(h264_reader::nal::pps::ParamSetId::from_u32(0).unwrap())
.ok_or_else(|| format_err!("No SPS 0"))?;
};
let mut sample_entry = Vec::with_capacity(256);
// This is a concatenation of the following boxes/classes.
// SampleEntry, ISO/IEC 14496-12 section 8.5.2.
let avc1_len_pos = sample_entry.len();
// length placeholder + type + reserved + data_reference_index = 1
sample_entry.extend_from_slice(b"\x00\x00\x00\x00avc1\x00\x00\x00\x00\x00\x00\x00\x01");
// VisualSampleEntry, ISO/IEC 14496-12 section 12.1.3.
sample_entry.extend_from_slice(&[0; 16]); // pre-defined + reserved
sample_entry.write_u16::<BigEndian>(width)?;
sample_entry.write_u16::<BigEndian>(height)?;
sample_entry.extend_from_slice(&[
0x00, 0x48, 0x00, 0x00, // horizresolution
0x00, 0x48, 0x00, 0x00, // vertresolution
0x00, 0x00, 0x00, 0x00, // reserved
0x00, 0x01, // frame count
0x00, 0x00, 0x00, 0x00, // compressorname
0x00, 0x00, 0x00, 0x00, //
0x00, 0x00, 0x00, 0x00, //
0x00, 0x00, 0x00, 0x00, //
0x00, 0x00, 0x00, 0x00, //
0x00, 0x00, 0x00, 0x00, //
0x00, 0x00, 0x00, 0x00, //
0x00, 0x00, 0x00, 0x00, //
0x00, 0x18, 0xff, 0xff, // depth + pre_defined
]);
// AVCSampleEntry, ISO/IEC 14496-15 section 5.3.4.1.
// AVCConfigurationBox, ISO/IEC 14496-15 section 5.3.4.1.
let avcc_len_pos = sample_entry.len();
sample_entry.extend_from_slice(b"\x00\x00\x00\x00avcC");
if let Some((sps, pps)) = raw_sps_and_pps {
// Create the AVCDecoderConfiguration, ISO/IEC 14496-15 section 5.2.4.1.
// The beginning of the AVCDecoderConfiguration takes a few values from
// the SPS (ISO/IEC 14496-10 section 7.3.2.1.1). One caveat: that section
// defines the syntax in terms of RBSP, not NAL. The difference is the
// escaping of 00 00 01 and 00 00 02; see notes about
// "emulation_prevention_three_byte" in ISO/IEC 14496-10 section 7.4.
// It looks like 00 is not a valid value of profile_idc, so this distinction
// shouldn't be relevant here. And ffmpeg seems to ignore it.
sample_entry.push(1); // configurationVersion
sample_entry.push(sps[1]); // profile_idc . AVCProfileIndication
sample_entry.push(sps[2]); // ...misc bits... . profile_compatibility
sample_entry.push(sps[3]); // level_idc . AVCLevelIndication
// Hardcode lengthSizeMinusOne to 3, matching TransformSampleData's 4-byte
// lengths.
sample_entry.push(0xff);
// Only support one SPS and PPS.
// ffmpeg's ff_isom_write_avcc has the same limitation, so it's probably
// fine. This next byte is a reserved 0b111 + a 5-bit # of SPSs (1).
sample_entry.push(0xe1);
sample_entry.write_u16::<BigEndian>(u16::try_from(sps.len())?)?;
sample_entry.extend_from_slice(sps);
sample_entry.push(1); // # of PPSs.
sample_entry.write_u16::<BigEndian>(u16::try_from(pps.len())?)?;
sample_entry.extend_from_slice(pps);
} else {
sample_entry.extend_from_slice(extradata);
};
// Fix up avc1 and avcC box lengths.
let cur_pos = sample_entry.len();
BigEndian::write_u32(
&mut sample_entry[avcc_len_pos..avcc_len_pos + 4],
u32::try_from(cur_pos - avcc_len_pos)?,
);
// PixelAspectRatioBox, ISO/IEC 14496-12 section 12.1.4.2.
// Write a PixelAspectRatioBox if necessary, as the sub streams can be be anamorphic.
let pasp = sps
.vui_parameters
.as_ref()
.and_then(|v| v.aspect_ratio_info.as_ref())
.and_then(|a| a.clone().get())
.unwrap_or_else(|| default_pixel_aspect_ratio(width, height));
if pasp != (1, 1) {
sample_entry.extend_from_slice(b"\x00\x00\x00\x10pasp"); // length + box name
sample_entry.write_u32::<BigEndian>(pasp.0.into())?;
sample_entry.write_u32::<BigEndian>(pasp.1.into())?;
}
let cur_pos = sample_entry.len();
BigEndian::write_u32(
&mut sample_entry[avc1_len_pos..avc1_len_pos + 4],
u32::try_from(cur_pos - avc1_len_pos)?,
);
let profile_idc = sample_entry[103];
let constraint_flags = sample_entry[104];
let level_idc = sample_entry[105];
let rfc6381_codec = format!(
"avc1.{:02x}{:02x}{:02x}",
profile_idc, constraint_flags, level_idc
);
Ok(ExtraData {
entry: db::VideoSampleEntryToInsert {
data: sample_entry,
rfc6381_codec,
width,
height,
pasp_h_spacing: pasp.0,
pasp_v_spacing: pasp.1,
},
need_transform,
})
}
}
/// Transforms sample data from Annex B format to AVC format. Should be called on samples iff
/// `ExtraData::need_transform` is true. Uses an out parameter `avc_sample` rather than a return
/// so that memory allocations can be reused from sample to sample.
pub fn transform_sample_data(annexb_sample: &[u8], avc_sample: &mut Vec<u8>) -> Result<(), Error> {
// See AVCParameterSamples, ISO/IEC 14496-15 section 5.3.2.
avc_sample.clear();
// The output will be about as long as the input. Annex B stop codes require at least three
// bytes; many seem to be four. The output lengths are exactly four.
avc_sample.reserve(annexb_sample.len() + 4);
decode_h264_annex_b(annexb_sample, |unit| {
// 4-byte length; this must match ParseExtraData's lengthSizeMinusOne == 3.
avc_sample.write_u32::<BigEndian>(unit.len() as u32)?; // length
avc_sample.extend_from_slice(unit);
Ok(())
})?;
Ok(())
}
#[cfg(test)]
mod tests {
use db::testutil;
#[rustfmt::skip]
const ANNEX_B_TEST_INPUT: [u8; 35] = [
0x00, 0x00, 0x00, 0x01, 0x67, 0x4d, 0x00, 0x1f,
0x9a, 0x66, 0x02, 0x80, 0x2d, 0xff, 0x35, 0x01,
0x01, 0x01, 0x40, 0x00, 0x00, 0xfa, 0x00, 0x00,
0x1d, 0x4c, 0x01, 0x00, 0x00, 0x00, 0x01, 0x68,
0xee, 0x3c, 0x80,
];
#[rustfmt::skip]
const AVC_DECODER_CONFIG_TEST_INPUT: [u8; 38] = [
0x01, 0x4d, 0x00, 0x1f, 0xff, 0xe1, 0x00, 0x17,
0x67, 0x4d, 0x00, 0x1f, 0x9a, 0x66, 0x02, 0x80,
0x2d, 0xff, 0x35, 0x01, 0x01, 0x01, 0x40, 0x00,
0x00, 0xfa, 0x00, 0x00, 0x1d, 0x4c, 0x01, 0x01,
0x00, 0x04, 0x68, 0xee, 0x3c, 0x80,
];
#[rustfmt::skip]
const TEST_OUTPUT: [u8; 132] = [
0x00, 0x00, 0x00, 0x84, 0x61, 0x76, 0x63, 0x31,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x05, 0x00, 0x02, 0xd0, 0x00, 0x48, 0x00, 0x00,
0x00, 0x48, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x18, 0xff, 0xff, 0x00, 0x00,
0x00, 0x2e, 0x61, 0x76, 0x63, 0x43, 0x01, 0x4d,
0x00, 0x1f, 0xff, 0xe1, 0x00, 0x17, 0x67, 0x4d,
0x00, 0x1f, 0x9a, 0x66, 0x02, 0x80, 0x2d, 0xff,
0x35, 0x01, 0x01, 0x01, 0x40, 0x00, 0x00, 0xfa,
0x00, 0x00, 0x1d, 0x4c, 0x01, 0x01, 0x00, 0x04,
0x68, 0xee, 0x3c, 0x80,
];
#[test]
fn test_decode() {
testutil::init();
let data = &ANNEX_B_TEST_INPUT;
let mut pieces = Vec::new();
super::decode_h264_annex_b(data, |p| {
pieces.push(p);
Ok(())
})
.unwrap();
assert_eq!(&pieces, &[&data[4..27], &data[31..]]);
}
#[test]
fn test_sample_entry_from_avc_decoder_config() {
testutil::init();
let e = super::ExtraData::parse(&AVC_DECODER_CONFIG_TEST_INPUT, 1280, 720).unwrap();
assert_eq!(&e.entry.data[..], &TEST_OUTPUT[..]);
assert_eq!(e.entry.width, 1280);
assert_eq!(e.entry.height, 720);
assert_eq!(e.entry.rfc6381_codec, "avc1.4d001f");
assert_eq!(e.need_transform, false);
}
#[test]
fn test_sample_entry_from_annex_b() {
testutil::init();
let e = super::ExtraData::parse(&ANNEX_B_TEST_INPUT, 1280, 720).unwrap();
assert_eq!(e.entry.width, 1280);
assert_eq!(e.entry.height, 720);
assert_eq!(e.entry.rfc6381_codec, "avc1.4d001f");
assert_eq!(e.need_transform, true);
}
#[test]
fn test_transform_sample_data() {
testutil::init();
#[rustfmt::skip]
const INPUT: [u8; 64] = [
0x00, 0x00, 0x00, 0x01, 0x67, 0x4d, 0x00, 0x1f,
0x9a, 0x66, 0x02, 0x80, 0x2d, 0xff, 0x35, 0x01,
0x01, 0x01, 0x40, 0x00, 0x00, 0xfa, 0x00, 0x00,
0x1d, 0x4c, 0x01,
0x00, 0x00, 0x00, 0x01, 0x68, 0xee, 0x3c, 0x80,
0x00, 0x00, 0x00, 0x01, 0x06, 0x06, 0x01, 0xc4,
0x80,
0x00, 0x00, 0x00, 0x01, 0x65, 0x88, 0x80, 0x10,
0x00, 0x08, 0x7f, 0x00, 0x5d, 0x27, 0xb5, 0xc1,
0xff, 0x8c, 0xd6, 0x35,
// (truncated)
];
#[rustfmt::skip]
const EXPECTED_OUTPUT: [u8; 64] = [
0x00, 0x00, 0x00, 0x17, 0x67, 0x4d, 0x00, 0x1f,
0x9a, 0x66, 0x02, 0x80, 0x2d, 0xff, 0x35, 0x01,
0x01, 0x01, 0x40, 0x00, 0x00, 0xfa, 0x00, 0x00,
0x1d, 0x4c, 0x01,
0x00, 0x00, 0x00, 0x04, 0x68, 0xee, 0x3c, 0x80,
0x00, 0x00, 0x00, 0x05, 0x06, 0x06, 0x01, 0xc4,
0x80,
0x00, 0x00, 0x00, 0x10, 0x65, 0x88, 0x80, 0x10,
0x00, 0x08, 0x7f, 0x00, 0x5d, 0x27, 0xb5, 0xc1,
0xff, 0x8c, 0xd6, 0x35,
];
let mut out = Vec::new();
super::transform_sample_data(&INPUT, &mut out).unwrap();
assert_eq!(&out[..], &EXPECTED_OUTPUT[..]);
}
#[test]
fn pixel_aspect_ratios() {
use super::default_pixel_aspect_ratio;
use num_rational::Ratio;
for &((w, h), _) in &super::PIXEL_ASPECT_RATIOS {
let (h_spacing, v_spacing) = default_pixel_aspect_ratio(w, h);
assert_eq!(Ratio::new(w * h_spacing, h * v_spacing), Ratio::new(16, 9));
// 90 or 270 degree rotation.
let (h_spacing, v_spacing) = default_pixel_aspect_ratio(h, w);
assert_eq!(Ratio::new(h * h_spacing, w * v_spacing), Ratio::new(9, 16));
}
}
}