Integrate AV-Metrics for calculation of metrics

For showing metrics in CLI use --metrics for showing all metrics and use --psnr for showing PNSR.

This commit introduces:
- Frame Metrics function to calculate
PSNR, PSNR-HVS, SSIM, MS-SSIM, CIEDE2000.
- Quality Metrics Structure to store all the calculated values
- Adding METRICS as an argument
- Updates --psnr calculation based on av-metrics
- Calculated metrics in a neat way
- Introduce metrics_cli as an additional parameter for
process_frame as parse_cli is an expensive function to be made in
encode_loop, to make it more efficient we have moved metrics_cli
enum as an argument to process_frame and making the calculation
in do_encode function.

Cargo.toml

@@ -31,6 +31,7 @@ binaries = [
     "fern",
     "console",
     "better-panic",
+    "av-metrics",
 ]
 default = ["binaries", "asm", "signal_support"]
 asm = ["nasm-rs", "cc"]
@@ -69,6 +70,7 @@ aom-sys = { version = "0.1.3", optional = true }
 scan_fmt = { version = "0.2.3", optional = true, default-features = false }
 ivf = { version = "0.1", path = "ivf/", optional = true }
 v_frame = { version = "0.1", path = "v_frame/" }
+av-metrics = { version = "0.4", optional = true }
 rayon = "1.0"
 toml = { version = "0.5", optional = true }
 arrayvec = "0.5"
@@ -104,6 +106,9 @@ version = "0.1.7"
 optional = true
 features = ["parallel"]
 
+[replace]
+"v_frame:0.1.0" = { path = "v_frame/" }
+
 [target.'cfg(unix)'.dependencies]
 signal-hook = { version = "0.1.9", optional = true }
 

src/api/config.rs

@@ -109,10 +109,6 @@ pub struct EncoderConfig {
   pub tiles: usize,
   /// Number of frames to read ahead for the RDO lookahead computation.
   pub rdo_lookahead_frames: usize,
-  /// If enabled, computes the PSNR values and stores them in [`Packet`].
-  ///
-  /// [`Packet`]: struct.Packet.html#structfield.psnr
-  pub show_psnr: bool,
 
   /// Settings which affect the enconding speed vs. quality trade-off.
   pub speed_settings: SpeedSettings,
@@ -171,7 +167,6 @@ impl EncoderConfig {
       tiles: 0,
       rdo_lookahead_frames: 40,
       speed_settings: SpeedSettings::from_preset(speed),
-      show_psnr: false,
     }
   }
 

src/api/internal.rs

@@ -16,7 +16,6 @@ use crate::dist::get_satd;
 use crate::encoder::*;
 use crate::frame::*;
 use crate::hawktracer::*;
-use crate::metrics::calculate_frame_psnr;
 use crate::partition::*;
 use crate::rate::RCState;
 use crate::rate::FRAME_NSUBTYPES;
@@ -351,18 +350,6 @@ impl<T: Pixel> ContextInner<T> {
     Ok(())
   }
 
-  fn get_frame(&self, input_frameno: u64) -> Arc<Frame<T>> {
-    // Clones only the arc, so low cost overhead
-    self
-      .frame_q
-      .get(&input_frameno)
-      .as_ref()
-      .unwrap()
-      .as_ref()
-      .unwrap()
-      .clone()
-  }
-
   /// Indicates whether more frames need to be read into the frame queue
   /// in order for frame queue lookahead to be full.
   fn needs_more_frame_q_lookahead(&self, input_frameno: u64) -> bool {
@@ -1101,15 +1088,13 @@ impl<T: Pixel> ContextInner<T> {
 
       let input_frameno = frame_data.fi.input_frameno;
       let frame_type = frame_data.fi.frame_type;
-      let bit_depth = frame_data.fi.sequence.bit_depth;
       let qp = frame_data.fi.base_q_idx;
       let enc_stats = frame_data.fs.enc_stats.clone();
       self.finalize_packet(
         rec,
         source,
         input_frameno,
         frame_type,
-        bit_depth,
         qp,
         enc_stats,
       )
@@ -1222,14 +1207,12 @@ impl<T: Pixel> ContextInner<T> {
       if fi.show_frame {
         let input_frameno = fi.input_frameno;
         let frame_type = fi.frame_type;
-        let bit_depth = fi.sequence.bit_depth;
         let qp = fi.base_q_idx;
         self.finalize_packet(
           rec,
           source,
           input_frameno,
           frame_type,
-          bit_depth,
           qp,
           enc_stats,
         )
@@ -1283,7 +1266,7 @@ impl<T: Pixel> ContextInner<T> {
 
   fn finalize_packet(
     &mut self, rec: Option<Arc<Frame<T>>>, source: Option<Arc<Frame<T>>>,
-    input_frameno: u64, frame_type: FrameType, bit_depth: usize, qp: u8,
+    input_frameno: u64, frame_type: FrameType, qp: u8,
     enc_stats: EncoderStats,
   ) -> Result<Packet<T>, EncoderStatus> {
     let data = self.packet_data.clone();
@@ -1292,25 +1275,8 @@ impl<T: Pixel> ContextInner<T> {
       return Err(EncoderStatus::Failure);
     }
 
-    let mut psnr = None;
-    if self.config.show_psnr {
-      if let Some(ref rec) = rec {
-        let original_frame = self.get_frame(input_frameno);
-        psnr = Some(calculate_frame_psnr(&*original_frame, rec, bit_depth));
-      }
-    }
-
     self.frames_processed += 1;
-    Ok(Packet {
-      data,
-      rec,
-      source,
-      input_frameno,
-      frame_type,
-      psnr,
-      qp,
-      enc_stats,
-    })
+    Ok(Packet { data, rec, source, input_frameno, frame_type, qp, enc_stats })
   }
 
   fn garbage_collect(&mut self, cur_input_frameno: u64) {

src/api/test.rs

@@ -1710,7 +1710,6 @@ fn log_q_exp_overflow() {
         non_square_partition: false,
         ..Default::default()
       },
-      show_psnr: false,
     },
     threads: 1,
   };
@@ -1778,7 +1777,6 @@ fn guess_frame_subtypes_assert() {
         non_square_partition: false,
         ..Default::default()
       },
-      show_psnr: false,
     },
     threads: 1,
   };

src/api/util.rs

@@ -170,8 +170,6 @@ pub struct Packet<T: Pixel> {
   pub input_frameno: u64,
   /// Type of the shown frame.
   pub frame_type: FrameType,
-  /// PSNR for Y, U, and V planes for the shown frame.
-  pub psnr: Option<(f64, f64, f64)>,
   /// QP selected for the frame.
   pub qp: u8,
   /// Block-level encoding stats for the frame

src/bin/common.rs

@@ -9,6 +9,7 @@
 
 use crate::error::*;
 use crate::muxer::{create_muxer, Muxer};
+use crate::stats::MetricsEnabled;
 use crate::{ColorPrimaries, MatrixCoefficients, TransferCharacteristics};
 use clap::{App, AppSettings, Arg, ArgMatches, Shell, SubCommand};
 use rav1e::prelude::*;
@@ -42,6 +43,7 @@ pub struct CliOptions {
   pub verbose: Verbose,
   pub benchmark: bool,
   pub threads: usize,
+  pub metrics_enabled: MetricsEnabled,
   pub pass1file_name: Option<String>,
   pub pass2file_name: Option<String>,
   pub save_config: Option<String>,
@@ -74,6 +76,8 @@ fn build_speed_long_help() -> String {
 }
 
 #[allow(unused_mut)]
+/// Only call this once at the start of the app,
+/// otherwise bad things will happen.
 pub fn parse_cli() -> Result<CliOptions, CliError> {
   let ver_short = version::short();
   let ver_long = version::full();
@@ -340,6 +344,11 @@ pub fn parse_cli() -> Result<CliOptions, CliError> {
         .help("Calculate and display PSNR metrics")
         .long("psnr")
     )
+    .arg(
+      Arg::with_name("METRICS")
+        .help("Calulate and display several metrics including PSNR, SSIM, CIEDE2000 etc")
+        .long("metrics")
+    )
     .arg(
       Arg::with_name("RECONSTRUCTION")
         .help("Outputs a Y4M file containing the output from the decoder")
@@ -448,6 +457,14 @@ pub fn parse_cli() -> Result<CliOptions, CliError> {
     Verbose::Normal
   };
 
+  let metrics_enabled = if matches.is_present("METRICS") {
+    MetricsEnabled::All
+  } else if matches.is_present("PSNR") {
+    MetricsEnabled::Psnr
+  } else {
+    MetricsEnabled::None
+  };
+
   Ok(CliOptions {
     io,
     enc,
@@ -456,6 +473,7 @@ pub fn parse_cli() -> Result<CliOptions, CliError> {
     // if a parameter has a default value.
     color_range_specified: matches.occurrences_of("PIXEL_RANGE") > 0,
     override_time_base: matches.is_present("FRAME_RATE"),
+    metrics_enabled,
     skip: matches.value_of("SKIP").unwrap().parse().unwrap(),
     benchmark: matches.is_present("BENCHMARK"),
     verbose,
@@ -615,7 +633,6 @@ fn parse_config(matches: &ArgMatches<'_>) -> Result<EncoderConfig, CliError> {
     .map(|reservior_frame_delay| reservior_frame_delay.parse().unwrap());
   cfg.rdo_lookahead_frames =
     matches.value_of("RDO_LOOKAHEAD_FRAMES").unwrap_or("40").parse().unwrap();
-  cfg.show_psnr = matches.is_present("PSNR");
   cfg.tune = matches.value_of("TUNE").unwrap().parse().unwrap();
 
   if cfg.tune == Tune::Psychovisual {

src/bin/rav1e.rs

@@ -100,6 +100,7 @@ fn process_frame<T: Pixel, D: Decoder>(
   pass1file: Option<&mut File>, pass2file: Option<&mut File>,
   buffer: &mut [u8], buf_pos: &mut usize,
   mut y4m_enc: Option<&mut y4m::Encoder<'_, Box<dyn Write>>>,
+  metrics_cli: MetricsEnabled,
 ) -> Result<Option<Vec<FrameSummary>>, CliError> {
   let y4m_details = source.input.get_video_details();
   let mut frame_summaries = Vec::new();
@@ -155,7 +156,12 @@ fn process_frame<T: Pixel, D: Decoder>(
       {
         write_y4m_frame(y4m_enc_uw, rec, y4m_details);
       }
-      frame_summaries.push(pkt.into());
+      frame_summaries.push(build_frame_summary(
+        pkt,
+        y4m_details.bit_depth,
+        y4m_details.chroma_sampling,
+        metrics_cli,
+      ));
     }
     Err(EncoderStatus::NeedMoreData) => {
       source.read_frame(ctx, y4m_details);
@@ -197,6 +203,7 @@ fn do_encode<T: Pixel, D: Decoder>(
   output: &mut dyn Muxer, source: &mut Source<D>,
   pass1file_name: Option<&String>, pass2file_name: Option<&String>,
   mut y4m_enc: Option<y4m::Encoder<'_, Box<dyn Write>>>,
+  metrics_enabled: MetricsEnabled,
 ) -> Result<(), CliError> {
   let mut ctx: Context<T> =
     cfg.new_context().map_err(|e| e.context("Invalid encoder settings"))?;
@@ -224,6 +231,7 @@ fn do_encode<T: Pixel, D: Decoder>(
     &mut buffer,
     &mut buf_pos,
     y4m_enc.as_mut(),
+    metrics_enabled,
   )? {
     if verbose != Verbose::Quiet {
       for frame in frame_info {
@@ -419,7 +427,7 @@ fn run() -> Result<(), error::CliError> {
   let progress = ProgressInfo::new(
     Rational { num: video_info.time_base.den, den: video_info.time_base.num },
     if cli.limit == 0 { None } else { Some(cli.limit) },
-    cfg.enc.show_psnr,
+    cli.metrics_enabled,
   );
 
   for _ in 0..cli.skip {
@@ -467,6 +475,7 @@ fn run() -> Result<(), error::CliError> {
       cli.pass1file_name.as_ref(),
       cli.pass2file_name.as_ref(),
       y4m_enc,
+      cli.metrics_enabled,
     )?
   } else {
     do_encode::<u16, y4m::Decoder<'_, Box<dyn Read>>>(
@@ -478,6 +487,7 @@ fn run() -> Result<(), error::CliError> {
       cli.pass1file_name.as_ref(),
       cli.pass2file_name.as_ref(),
       y4m_enc,
+      cli.metrics_enabled,
     )?
   }
   if cli.benchmark {