Merge pull request #296 from dreamaniajp/master

Add support for MPEG-4 Part 2 video codec (ISO 14496-2).

Author: baumanj

mp4parse/Cargo.toml

@@ -44,6 +44,7 @@ criterion = "0.3"
 3gpp = []
 meta-xml = []
 unstable-api = []
+mp4v = []
 
 [[bench]]
 name = "avif_benchmark"

mp4parse/src/lib.rs

@@ -407,6 +407,8 @@ pub struct ES_Descriptor {
     pub extended_audio_object_type: Option<u16>,
     pub audio_sample_rate: Option<u32>,
     pub audio_channel_count: Option<u16>,
+    #[cfg(feature = "mp4v")]
+    pub video_codec: CodecType,
     pub codec_esds: TryVec<u8>,
     pub decoder_specific_data: TryVec<u8>, // Data in DECODER_SPECIFIC_TAG
 }
@@ -3725,6 +3727,14 @@ fn get_audio_object_type(bit_reader: &mut BitReader) -> Result<u16> {
 
 /// See MPEG-4 Systems (ISO 14496-1:2010) § 7.2.6.7 and probably 14496-3 somewhere?
 fn read_ds_descriptor(data: &[u8], esds: &mut ES_Descriptor) -> Result<()> {
+    #[cfg(feature = "mp4v")]
+    // Check if we are in a Visual esda Box.
+    if esds.video_codec != CodecType::Unknown {
+        esds.decoder_specific_data.extend_from_slice(data)?;
+        return Ok(());
+    }
+
+    // We are in an Audio esda Box.
     let frequency_table = vec![
         (0x0, 96000),
         (0x1, 88200),
@@ -3899,6 +3909,14 @@ fn read_dc_descriptor(data: &[u8], esds: &mut ES_Descriptor) -> Result<()> {
     let des = &mut Cursor::new(data);
     let object_profile = des.read_u8()?;
 
+    #[cfg(feature = "mp4v")]
+    {
+        esds.video_codec = match object_profile {
+            0x20..=0x24 => CodecType::MP4V,
+            _ => CodecType::Unknown,
+        };
+    }
+
     // Skip uninteresting fields.
     skip(des, 12)?;
 
@@ -4214,16 +4232,27 @@ fn read_video_sample_entry<T: Read>(src: &mut BMFFBox<T>) -> Result<SampleEntry>
                 if name != BoxType::MP4VideoSampleEntry || codec_specific.is_some() {
                     return Err(Error::InvalidData("malformed video sample entry"));
                 }
-                let (_, _) = read_fullbox_extra(&mut b.content)?;
-                // Subtract 4 extra to offset the members of fullbox not
-                // accounted for in head.offset
-                let esds_size = b
-                    .head
-                    .size
-                    .checked_sub(b.head.offset + 4)
-                    .expect("offset invalid");
-                let esds = read_buf(&mut b.content, esds_size)?;
-                codec_specific = Some(VideoCodecSpecific::ESDSConfig(esds));
+                #[cfg(not(feature = "mp4v"))]
+                {
+                    let (_, _) = read_fullbox_extra(&mut b.content)?;
+                    // Subtract 4 extra to offset the members of fullbox not
+                    // accounted for in head.offset
+                    let esds_size = b
+                        .head
+                        .size
+                        .checked_sub(b.head.offset + 4)
+                        .expect("offset invalid");
+                    let esds = read_buf(&mut b.content, esds_size)?;
+                    codec_specific = Some(VideoCodecSpecific::ESDSConfig(esds));
+                }
+                #[cfg(feature = "mp4v")]
+                {
+                    // Read ES_Descriptor inside an esds box.
+                    // See ISOBMFF (ISO 14496-1:2010 §7.2.6.5)
+                    let esds = read_esds(&mut b)?;
+                    codec_specific =
+                        Some(VideoCodecSpecific::ESDSConfig(esds.decoder_specific_data));
+                }
             }
             BoxType::ProtectionSchemeInfoBox => {
                 if name != BoxType::ProtectedVisualSampleEntry {

mp4parse/src/tests.rs

@@ -1039,8 +1039,10 @@ fn read_stsd_mp4v() {
         0x2e, 0xa6, 0x60, 0x16, 0xf4, 0x01, 0xf4, 0x24, 0xc8, 0x01, 0xe5, 0x16, 0x84, 0x3c, 0x14,
         0x63, 0x06, 0x01, 0x02,
     ];
-
+    #[cfg(not(feature = "mp4v"))]
     let esds_specific_data = &mp4v[90..];
+    #[cfg(feature = "mp4v")]
+    let esds_specific_data = &mp4v[112..151];
     println!("esds_specific_data {:?}", esds_specific_data);
 
     let mut stream = make_box(BoxSize::Auto, b"mp4v", |s| s.append_bytes(mp4v.as_slice()));

mp4parse/tests/bbb_sunflower_QCIF_30fps_mp4v_noaudio_1f.mp4

@@ -79,6 +79,10 @@ static AUDIO_AMRNB_3GP: &str = "tests/amr_nb_1f.3gp";
 // "ffmpeg -i [input file] -f 3gp -acodec amr_wb -ar 16000 -ac 1 -frames:a 1 -vn output.3gp"
 #[cfg(feature = "3gpp")]
 static AUDIO_AMRWB_3GP: &str = "tests/amr_wb_1f.3gp";
+#[cfg(feature = "mp4v")]
+// The 1 frame mp4v mp4 file can be generated by ffmpeg with command
+// "ffmpeg -i [input file] -f mp4 -c:v mpeg4 -vf scale=176x144 -frames:v 1 -an output.mp4"
+static VIDEO_MP4V_MP4: &str = "tests/bbb_sunflower_QCIF_30fps_mp4v_noaudio_1f.mp4";
 
 // Adapted from https://github.com/GuillaumeGomez/audio-video-metadata/blob/9dff40f565af71d5502e03a2e78ae63df95cfd40/src/metadata.rs#L53
 #[test]
@@ -1054,3 +1058,30 @@ fn public_audio_amrwb() {
         };
     }
 }
+
+#[test]
+#[cfg(feature = "mp4v")]
+fn public_video_mp4v() {
+    let mut fd = File::open(VIDEO_MP4V_MP4).expect("Unknown file");
+    let mut buf = Vec::new();
+    fd.read_to_end(&mut buf).expect("File error");
+
+    let mut c = Cursor::new(&buf);
+    let context = mp4::read_mp4(&mut c).expect("read_mp4 failed");
+    for track in context.tracks {
+        let stsd = track.stsd.expect("expected an stsd");
+        let v = match stsd.descriptions.first().expect("expected a SampleEntry") {
+            mp4::SampleEntry::Video(ref v) => v,
+            _ => panic!("expected a VideoSampleEntry"),
+        };
+        assert_eq!(v.codec_type, mp4::CodecType::MP4V);
+        assert_eq!(v.width, 176);
+        assert_eq!(v.height, 144);
+        let _codec_specific = match v.codec_specific {
+            mp4::VideoCodecSpecific::ESDSConfig(_) => true,
+            _ => {
+                panic!("expected a ESDSConfig",);
+            }
+        };
+    }
+}

mp4parse/tests/public.rs

@@ -36,3 +36,4 @@ env_logger = "0.8"
 [features]
 3gpp = ["mp4parse/3gpp"]
 meta-xml = ["mp4parse/meta-xml"]
+mp4v = ["mp4parse/mp4v"]

mp4parse_capi/Cargo.toml

@@ -973,6 +973,9 @@ fn mp4parse_get_track_video_info_safe(
             VideoCodecSpecific::AV1Config(_) => Mp4parseCodec::Av1,
             VideoCodecSpecific::AVCConfig(_) => Mp4parseCodec::Avc,
             VideoCodecSpecific::H263Config(_) => Mp4parseCodec::H263,
+            #[cfg(feature = "mp4v")]
+            VideoCodecSpecific::ESDSConfig(_) => Mp4parseCodec::Mp4v,
+            #[cfg(not(feature = "mp4v"))]
             VideoCodecSpecific::ESDSConfig(_) =>
             // MP4V (14496-2) video is unsupported.
             {