Implement support for encoding HDR10+ from JSON metadata file

Author: quietvoid

Cargo.toml

@@ -105,6 +105,7 @@ new_debug_unreachable = "1.0.4"
 once_cell = "1.13.0"
 av1-grain = { version = "0.1.1", features = ["serialize"] }
 serde-big-array = { version = "0.4.1", optional = true }
+hdr10plus = { version = "1.1.1", features = ["json"] }
 
 [dependencies.image]
 version = "0.23"

src/api/config/encoder.rs

@@ -15,6 +15,9 @@ use crate::api::{Rational, SpeedSettings};
 use crate::encoder::Tune;
 use crate::serialize::{Deserialize, Serialize};
 
+#[cfg(feature = "unstable")]
+use std::collections::BTreeMap;
+
 use std::fmt;
 
 // We add 1 to rdo_lookahead_frames in a bunch of places.
@@ -85,6 +88,9 @@ pub struct EncoderConfig {
   pub tune: Tune,
   /// Parameters for grain synthesis.
   pub film_grain_params: Option<Vec<GrainTableSegment>>,
+  /// HDR10+ T.35 metadata payload map, by frame index.
+  #[cfg(feature = "unstable")]
+  pub hdr10plus_payloads: Option<BTreeMap<u64, Vec<u8>>>,
   /// Number of tiles horizontally. Must be a power of two.
   ///
   /// Overridden by [`tiles`], if present.
@@ -162,6 +168,8 @@ impl EncoderConfig {
       bitrate: 0,
       tune: Tune::default(),
       film_grain_params: None,
+      #[cfg(feature = "unstable")]
+      hdr10plus_payloads: None,
       tile_cols: 0,
       tile_rows: 0,
       tiles: 0,

src/api/internal.rs

@@ -349,14 +349,35 @@ impl<T: Pixel> ContextInner<T> {
     }
     self.frame_q.insert(input_frameno, frame);
 
+    #[cfg(feature = "unstable")]
+    // Update T.35 metadata from encoder config
+    let maybe_updated_t35_metadata = self.get_maybe_updated_t35_metadata(
+      input_frameno,
+      params.as_ref().map(|params| params.t35_metadata.as_ref()),
+    );
+
     if let Some(params) = params {
       if params.frame_type_override == FrameTypeOverride::Key {
         self.keyframes_forced.insert(input_frameno);
       }
       if let Some(op) = params.opaque {
         self.opaque_q.insert(input_frameno, op);
       }
+
+      #[cfg(not(feature = "unstable"))]
       self.t35_q.insert(input_frameno, params.t35_metadata);
+
+      #[cfg(feature = "unstable")]
+      if let Some(new_t35_metadata) = maybe_updated_t35_metadata {
+        self.t35_q.insert(input_frameno, new_t35_metadata.into_boxed_slice());
+      } else {
+        self.t35_q.insert(input_frameno, params.t35_metadata);
+      }
+    } else {
+      #[cfg(feature = "unstable")]
+      if let Some(new_t35_metadata) = maybe_updated_t35_metadata {
+        self.t35_q.insert(input_frameno, new_t35_metadata.into_boxed_slice());
+      }
     }
 
     if !self.needs_more_frame_q_lookahead(self.next_lookahead_frame) {
@@ -1688,4 +1709,59 @@ impl<T: Pixel> ContextInner<T> {
       (prev_keyframe_nframes, prev_keyframe_ntus)
     }
   }
+
+  #[cfg(feature = "unstable")]
+  /// Updates the T.35 metadata to be added to the frame.
+  ///   The existing T.35 array may come from `FrameParameters`.
+  /// New metadata is added from the encoder config:
+  ///   - HDR10+, ST2094-40 in `EncoderConfig.hdr10plus_payloads`
+  ///
+  /// Returns an `Option`, where `None` means the T.35 metadata is unchanged.
+  /// Otherwise, the updated T.35 metadata array is returned.
+  fn get_maybe_updated_t35_metadata(
+    &self, input_frameno: u64, maybe_existing_t35_metadata: Option<&[T35]>,
+  ) -> Option<Vec<T35>> {
+    let hdr10plus_payload = self
+      .config
+      .hdr10plus_payloads
+      .as_ref()
+      .and_then(|list| list.get(&input_frameno));
+
+    let update_t35_metadata = hdr10plus_payload.is_some();
+
+    let mut new_t35_metadata = if update_t35_metadata {
+      Some(
+        maybe_existing_t35_metadata.map_or_else(Vec::new, |t35| t35.to_vec()),
+      )
+    } else {
+      None
+    };
+
+    if let Some(list) = new_t35_metadata.as_mut() {
+      // HDR10+, ST2094-40
+      if let Some(payload) = hdr10plus_payload {
+        // FIXME: Make const
+        let st2094_40_needle = &[
+          0x00, 0x03C, // Samsung Electronics America
+          0x00, 0x01, // ST-2094-40
+          0x04, // application_identifier = 4
+          0x01, // application_mode =1
+        ];
+
+        let has_existing_hdr10plus_meta = list.iter().any(|t35| {
+          t35.country_code == 0xB5 && t35.data.starts_with(st2094_40_needle)
+        });
+
+        if !has_existing_hdr10plus_meta {
+          list.push(T35 {
+            country_code: 0xB5,
+            country_code_extension_byte: 0x00,
+            data: payload.clone().into_boxed_slice(),
+          });
+        }
+      }
+    }
+
+    new_t35_metadata
+  }
 }

src/api/test.rs

@@ -2128,6 +2128,8 @@ fn log_q_exp_overflow() {
     bitrate: 1,
     tune: Tune::Psychovisual,
     film_grain_params: None,
+    #[cfg(feature = "unstable")]
+    hdr10plus_payloads: None,
     tile_cols: 0,
     tile_rows: 0,
     tiles: 0,
@@ -2205,6 +2207,8 @@ fn guess_frame_subtypes_assert() {
     bitrate: 16384,
     tune: Tune::Psychovisual,
     film_grain_params: None,
+    #[cfg(feature = "unstable")]
+    hdr10plus_payloads: None,
     tile_cols: 0,
     tile_rows: 0,
     tiles: 0,

src/bin/common.rs

@@ -17,6 +17,9 @@ use once_cell::sync::Lazy;
 use rav1e::prelude::*;
 use scan_fmt::scan_fmt;
 
+#[cfg(feature = "unstable")]
+use std::collections::BTreeMap;
+
 use std::fs::File;
 use std::io;
 use std::io::prelude::*;
@@ -203,6 +206,15 @@ pub struct CliOptions {
     help_heading = "ENCODE SETTINGS"
   )]
   pub film_grain_table: Option<PathBuf>,
+  /// Uses a HDR10+ metadata JSON file to add as T.35 metadata to the encode.
+  #[cfg(feature = "unstable")]
+  #[clap(
+    long,
+    alias = "dhdr10-info",
+    value_parser,
+    help_heading = "ENCODE SETTINGS"
+  )]
+  pub hdr10plus_json: Option<PathBuf>,
 
   /// Pixel range
   #[clap(long, value_parser, help_heading = "VIDEO METADATA")]
@@ -688,11 +700,42 @@ fn parse_config(matches: &CliOptions) -> Result<EncoderConfig, CliError> {
       .expect("Failed to read film grain table file");
     let table = av1_grain::parse_grain_table(&contents)
       .expect("Failed to parse film grain table");
+
     if !table.is_empty() {
       cfg.film_grain_params = Some(table);
     }
   }
 
+  #[cfg(feature = "unstable")]
+  if let Some(json_file) = matches.hdr10plus_json.as_ref() {
+    let contents = std::fs::read_to_string(json_file)
+      .expect("Failed to read HDR10+ metadata file");
+    let metadata_root =
+      hdr10plus::metadata_json::MetadataJsonRoot::parse(&contents)
+        .expect("Failed to parse HDR10+ metadata");
+
+    let payloads: BTreeMap<u64, Vec<u8>> = metadata_root
+      .scene_info
+      .iter()
+      .filter_map(|meta| {
+        hdr10plus::metadata::Hdr10PlusMetadata::try_from(meta)
+          .and_then(|meta| meta.encode(true))
+          .ok()
+          .map(|mut bytes| {
+            // Serialized with country code, which shouldn't be present
+            bytes.remove(0);
+            bytes
+          })
+      })
+      .zip(0u64..)
+      .map(|(payload, frame_no)| (frame_no, payload))
+      .collect();
+
+    if !payloads.is_empty() {
+      cfg.hdr10plus_payloads = Some(payloads);
+    }
+  }
+
   if let Some(frame_rate) = matches.frame_rate {
     cfg.time_base = Rational::new(matches.time_scale, frame_rate);
   }