Implement support for encoding Dolby Vision from RPU file

Cargo.toml

@@ -108,6 +108,7 @@ new_debug_unreachable = "1.0.4"
 once_cell = "1.18.0"
 av1-grain = { version = "0.2.2", features = ["serialize"] }
 serde-big-array = { version = "0.5.1", optional = true }
+dolby_vision = { version = "3.2.0" }
 
 [dependencies.image]
 version = "0.24.6"

src/api/internal.rs

@@ -520,12 +520,6 @@ impl<T: Pixel> ContextInner<T> {
       return Err(EncoderStatus::NeedMoreData);
     }
 
-    let t35_metadata = if let Some(t35) = self.t35_q.remove(&input_frameno) {
-      t35
-    } else {
-      Box::new([])
-    };
-
     if output_frameno_in_gop > 0 {
       let next_keyframe_input_frameno = self.next_keyframe_input_frameno(
         self.gop_input_frameno_start[&output_frameno],
@@ -554,6 +548,17 @@ impl<T: Pixel> ContextInner<T> {
           *self.gop_input_frameno_start.get_mut(&output_frameno).unwrap() =
             next_keyframe_input_frameno;
         } else {
+          let input_frameno = self.inter_cfg.get_input_frameno(
+            output_frameno_in_gop,
+            self.gop_input_frameno_start[&output_frameno],
+          );
+          let t35_metadata =
+            if let Some(t35) = self.t35_q.remove(&input_frameno) {
+              t35
+            } else {
+              Box::new([])
+            };
+
           let fi = FrameInvariants::new_inter_frame(
             self.get_previous_coded_fi(output_frameno),
             &self.inter_cfg,
@@ -592,6 +597,12 @@ impl<T: Pixel> ContextInner<T> {
     let output_frameno_in_gop =
       output_frameno - self.gop_output_frameno_start[&output_frameno];
     if output_frameno_in_gop == 0 {
+      let t35_metadata = if let Some(t35) = self.t35_q.remove(&input_frameno) {
+        t35
+      } else {
+        Box::new([])
+      };
+
       let fi = FrameInvariants::new_key_frame(
         self.config.clone(),
         self.seq.clone(),
@@ -600,6 +611,16 @@ impl<T: Pixel> ContextInner<T> {
       );
       Ok(Some(fi))
     } else {
+      let input_frameno = self.inter_cfg.get_input_frameno(
+        output_frameno_in_gop,
+        self.gop_input_frameno_start[&output_frameno],
+      );
+      let t35_metadata = if let Some(t35) = self.t35_q.remove(&input_frameno) {
+        t35
+      } else {
+        Box::new([])
+      };
+
       let next_keyframe_input_frameno = self.next_keyframe_input_frameno(
         self.gop_input_frameno_start[&output_frameno],
         false,

src/api/util.rs

@@ -8,6 +8,7 @@
 // PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 #![deny(missing_docs)]
 
+use crate::encoder::FrameInvariants;
 use crate::frame::*;
 use crate::serialize::{Deserialize, Serialize};
 use crate::stats::EncoderStats;
@@ -137,6 +138,12 @@ impl fmt::Display for FrameType {
   }
 }
 
+/// Dolby Vision T.35 metadata payload expected prefix.
+pub const T35_DOVI_PAYLOAD_PREFIX: &[u8] = &[
+  0x00, 0x03B, // Dolby
+  0x00, 0x00, 0x08, 0x00, 0x37, 0xCD, 0x08,
+];
+
 /// A single T.35 metadata packet.
 #[derive(Clone, Debug, Default)]
 pub struct T35 {
@@ -299,3 +306,24 @@ impl<T: Pixel> IntoFrame<T> for (Frame<T>, Option<FrameParameters>) {
     (Some(Arc::new(self.0)), self.1)
   }
 }
+
+impl T35 {
+  /// Whether the T.35 metadata is Dolby Vision Metadata.
+  pub fn is_dovi_metadata(&self) -> bool {
+    self.country_code == 0xB5 && self.data.starts_with(T35_DOVI_PAYLOAD_PREFIX)
+  }
+
+  /// Returns true if the T35 metadata can be added to the frame
+  pub fn is_valid_placement<T: Pixel>(&self, fi: &FrameInvariants<T>) -> bool {
+    if self.is_dovi_metadata() {
+      return fi.show_frame || fi.is_show_existing_frame();
+    }
+
+    true
+  }
+
+  /// Metadata that can be encoded with SEF frames
+  pub fn can_be_added_to_show_existing_frame(&self) -> bool {
+    self.is_dovi_metadata()
+  }
+}

src/bin/common.rs

@@ -18,6 +18,8 @@ use rav1e::prelude::*;
 use scan_fmt::scan_fmt;
 
 use rav1e::config::CpuFeatureLevel;
+
+use std::collections::BTreeMap;
 use std::fs::File;
 use std::io;
 use std::io::prelude::*;
@@ -195,6 +197,15 @@ pub struct CliOptions {
     help_heading = "ENCODE SETTINGS"
   )]
   pub film_grain_table: Option<PathBuf>,
+  /// Uses a Dolby Vision RPU file to add as T.35 metadata to the encode.
+  /// The RPU must be in the same format as for x265
+  #[clap(
+    long,
+    alias = "dolby-vision-rpu",
+    value_parser,
+    help_heading = "ENCODE SETTINGS"
+  )]
+  pub dovi_rpu: Option<PathBuf>,
 
   /// Pixel range
   #[clap(long, value_parser, help_heading = "VIDEO METADATA")]
@@ -339,6 +350,7 @@ pub struct ParsedCliOptions {
   pub photon_noise: u8,
   #[cfg(feature = "unstable")]
   pub slots: usize,
+  pub dovi_payloads: Option<BTreeMap<u64, T35>>,
 }
 
 #[cfg(feature = "serialize")]
@@ -466,6 +478,35 @@ pub fn parse_cli() -> Result<ParsedCliOptions, CliError> {
     panic!("A limit cannot be set above 1 in still picture mode");
   }
 
+  let dovi_payloads = if let Some(rpu_file) = matches.dovi_rpu.as_ref() {
+    let rpus = dolby_vision::rpu::utils::parse_rpu_file(rpu_file)
+      .expect("Failed to read Dolby Vision RPU file");
+
+    let payloads: BTreeMap<u64, T35> = rpus
+      .iter()
+      .filter_map(|rpu| {
+        rpu
+          .write_av1_rpu_metadata_obu_t35_payload()
+          .map(|payload| T35 {
+            country_code: 0xB5,
+            country_code_extension_byte: 0x00,
+            data: payload.into_boxed_slice(),
+          })
+          .ok()
+      })
+      .zip(0u64..)
+      .map(|(payload, frame_no)| (frame_no, payload))
+      .collect();
+
+    if !payloads.is_empty() {
+      Some(payloads)
+    } else {
+      None
+    }
+  } else {
+    None
+  };
+
   #[cfg(feature = "unstable")]
   let slots = matches.slots;
 
@@ -484,6 +525,7 @@ pub fn parse_cli() -> Result<ParsedCliOptions, CliError> {
     pass2file_name: matches.second_pass.clone(),
     save_config: save_config_path,
     photon_noise: matches.photon_noise,
+    dovi_payloads,
     #[cfg(feature = "unstable")]
     slots,
   })