aboutsummaryrefslogtreecommitdiff
path: root/src/Jellyfin.MediaEncoding.Keyframes
diff options
context:
space:
mode:
authorcvium <clausvium@gmail.com>2021-09-23 15:29:12 +0200
committercvium <clausvium@gmail.com>2021-09-23 15:29:12 +0200
commit9c15f96e12a0d48a70cbca8380bf78a4f2512b03 (patch)
tree068bc87052c9554afa788bcafd6022e7545f8189 /src/Jellyfin.MediaEncoding.Keyframes
parent1ebd3c9ac33ab99813307728ad6efbf53a667d4e (diff)
Add first draft of keyframe extraction for Matroska
Diffstat (limited to 'src/Jellyfin.MediaEncoding.Keyframes')
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/FfProbe/FfProbeKeyframeExtractor.cs10
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/FfTool/FfToolKeyframeExtractor.cs10
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/Jellyfin.MediaEncoding.Keyframes.csproj24
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/KeyframeData.cs28
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/KeyframeExtractor.cs56
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/Matroska/Extensions/EbmlReaderExtensions.cs181
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaConstants.cs31
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaKeyframeExtractor.cs76
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/Info.cs29
-rw-r--r--src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/SeekHead.cs36
10 files changed, 481 insertions, 0 deletions
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/FfProbe/FfProbeKeyframeExtractor.cs b/src/Jellyfin.MediaEncoding.Keyframes/FfProbe/FfProbeKeyframeExtractor.cs
new file mode 100644
index 000000000..249608ef9
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/FfProbe/FfProbeKeyframeExtractor.cs
@@ -0,0 +1,10 @@
+using System;
+
+namespace Jellyfin.MediaEncoding.Keyframes.FfProbe
+{
+ public static class FfProbeKeyframeExtractor
+ {
+ // TODO
+ public static KeyframeData GetKeyframeData(string ffProbePath, string filePath) => throw new NotImplementedException();
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/FfTool/FfToolKeyframeExtractor.cs b/src/Jellyfin.MediaEncoding.Keyframes/FfTool/FfToolKeyframeExtractor.cs
new file mode 100644
index 000000000..89c149ff4
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/FfTool/FfToolKeyframeExtractor.cs
@@ -0,0 +1,10 @@
+using System;
+
+namespace Jellyfin.MediaEncoding.Keyframes.FfTool
+{
+ public static class FfToolKeyframeExtractor
+ {
+ // TODO
+ public static KeyframeData GetKeyframeData(string ffProbePath, string filePath) => throw new NotImplementedException();
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/Jellyfin.MediaEncoding.Keyframes.csproj b/src/Jellyfin.MediaEncoding.Keyframes/Jellyfin.MediaEncoding.Keyframes.csproj
new file mode 100644
index 000000000..7a984658b
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/Jellyfin.MediaEncoding.Keyframes.csproj
@@ -0,0 +1,24 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+ <PropertyGroup>
+ <TargetFramework>net5.0</TargetFramework>
+ <RootNamespace>Jellyfin.MediaEncoding.Keyframes</RootNamespace>
+ </PropertyGroup>
+
+ <ItemGroup>
+ <PackageReference Include="NEbml" Version="0.11.0" />
+ </ItemGroup>
+
+ <!-- Code Analyzers-->
+ <ItemGroup Condition=" '$(Configuration)' == 'Debug' ">
+ <PackageReference Include="SerilogAnalyzer" Version="0.15.0" PrivateAssets="All" />
+ <PackageReference Include="StyleCop.Analyzers" Version="1.1.118" PrivateAssets="All" />
+ <PackageReference Include="SmartAnalyzers.MultithreadingAnalyzer" Version="1.1.31" PrivateAssets="All" />
+ </ItemGroup>
+ <ItemGroup>
+ <Reference Include="Microsoft.Extensions.Logging.Abstractions, Version=5.0.0.0, Culture=neutral, PublicKeyToken=adb9793829ddae60">
+ <HintPath>..\..\..\..\..\..\Program Files\dotnet\packs\Microsoft.AspNetCore.App.Ref\5.0.0\ref\net5.0\Microsoft.Extensions.Logging.Abstractions.dll</HintPath>
+ </Reference>
+ </ItemGroup>
+
+</Project>
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/KeyframeData.cs b/src/Jellyfin.MediaEncoding.Keyframes/KeyframeData.cs
new file mode 100644
index 000000000..3122f827c
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/KeyframeData.cs
@@ -0,0 +1,28 @@
+using System.Collections.Generic;
+
+namespace Jellyfin.MediaEncoding.Keyframes
+{
+ public class KeyframeData
+ {
+ /// <summary>
+ /// Initializes a new instance of the <see cref="KeyframeData"/> class.
+ /// </summary>
+ /// <param name="totalDuration">The total duration of the video stream in ticks.</param>
+ /// <param name="keyframeTicks">The video keyframes in ticks.</param>
+ public KeyframeData(long totalDuration, IReadOnlyList<long> keyframeTicks)
+ {
+ TotalDuration = totalDuration;
+ KeyframeTicks = keyframeTicks;
+ }
+
+ /// <summary>
+ /// Gets the total duration of the stream in ticks.
+ /// </summary>
+ public long TotalDuration { get; }
+
+ /// <summary>
+ /// Gets the keyframes in ticks.
+ /// </summary>
+ public IReadOnlyList<long> KeyframeTicks { get; }
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/KeyframeExtractor.cs b/src/Jellyfin.MediaEncoding.Keyframes/KeyframeExtractor.cs
new file mode 100644
index 000000000..2ee6b43e6
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/KeyframeExtractor.cs
@@ -0,0 +1,56 @@
+using System;
+using System.IO;
+using Jellyfin.MediaEncoding.Keyframes.FfProbe;
+using Jellyfin.MediaEncoding.Keyframes.FfTool;
+using Jellyfin.MediaEncoding.Keyframes.Matroska;
+using Microsoft.Extensions.Logging;
+
+namespace Jellyfin.MediaEncoding.Keyframes
+{
+ /// <summary>
+ /// Manager class for the set of keyframe extractors.
+ /// </summary>
+ public class KeyframeExtractor
+ {
+ private readonly ILogger<KeyframeExtractor> _logger;
+
+ /// <summary>
+ /// Initializes a new instance of the <see cref="KeyframeExtractor"/> class.
+ /// </summary>
+ /// <param name="logger">An instance of the <see cref="ILogger{KeyframeExtractor}"/> interface.</param>
+ public KeyframeExtractor(ILogger<KeyframeExtractor> logger)
+ {
+ _logger = logger;
+ }
+
+ /// <summary>
+ /// Extracts the keyframe positions from a video file.
+ /// </summary>
+ /// <param name="filePath">Absolute file path to the media file.</param>
+ /// <param name="ffProbePath">Absolute file path to the ffprobe executable.</param>
+ /// <param name="ffToolPath">Absolute file path to the fftool executable.</param>
+ /// <returns></returns>
+ public KeyframeData GetKeyframeData(string filePath, string ffProbePath, string ffToolPath)
+ {
+ var extension = Path.GetExtension(filePath);
+ if (string.Equals(extension, ".mkv", StringComparison.OrdinalIgnoreCase))
+ {
+ try
+ {
+ return MatroskaKeyframeExtractor.GetKeyframeData(filePath);
+ }
+ catch (InvalidOperationException ex)
+ {
+ _logger.LogError(ex, "{MatroskaKeyframeExtractor} failed to extract keyframes", nameof(MatroskaKeyframeExtractor));
+ }
+ }
+
+ if (!string.IsNullOrEmpty(ffToolPath))
+ {
+ return FfToolKeyframeExtractor.GetKeyframeData(ffToolPath, filePath);
+ }
+
+ return FfProbeKeyframeExtractor.GetKeyframeData(ffProbePath, filePath);
+ }
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Extensions/EbmlReaderExtensions.cs b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Extensions/EbmlReaderExtensions.cs
new file mode 100644
index 000000000..0de0f996c
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Extensions/EbmlReaderExtensions.cs
@@ -0,0 +1,181 @@
+using System;
+using Jellyfin.MediaEncoding.Keyframes.Matroska.Models;
+using NEbml.Core;
+
+namespace Jellyfin.MediaEncoding.Keyframes.Matroska.Extensions
+{
+ /// <summary>
+ /// Extension methods for the <see cref="EbmlReader"/> class.
+ /// </summary>
+ internal static class EbmlReaderExtensions
+ {
+ /// <summary>
+ /// Traverses the current container to find the element with <paramref name="identifier"/> identifier.
+ /// </summary>
+ /// <param name="reader">An instance of <see cref="EbmlReader"/>.</param>
+ /// <param name="identifier">The element identifier.</param>
+ /// <returns>A value indicating whether the element was found.</returns>
+ internal static bool FindElement(this EbmlReader reader, ulong identifier)
+ {
+ while (reader.ReadNext())
+ {
+ if (reader.ElementId.EncodedValue == identifier)
+ {
+ return true;
+ }
+ }
+
+ return false;
+ }
+
+ /// <summary>
+ /// Reads the current position in the file as an unsigned integer converted from binary.
+ /// </summary>
+ /// <param name="reader">An instance of <see cref="EbmlReader"/>.</param>
+ /// <returns>The unsigned integer.</returns>
+ internal static uint ReadUIntFromBinary(this EbmlReader reader)
+ {
+ var buffer = new byte[4];
+ reader.ReadBinary(buffer, 0, 4);
+ if (BitConverter.IsLittleEndian)
+ {
+ Array.Reverse(buffer);
+ }
+
+ return BitConverter.ToUInt32(buffer);
+ }
+
+ /// <summary>
+ /// Reads from the start of the file to retrieve the SeekHead segment.
+ /// </summary>
+ /// <param name="reader">An instance of <see cref="EbmlReader"/>.</param>
+ /// <returns>Instance of <see cref="SeekHead"/></returns>
+ internal static SeekHead ReadSeekHead(this EbmlReader reader)
+ {
+ reader = reader ?? throw new ArgumentNullException(nameof(reader));
+
+ if (reader.ElementPosition != 0)
+ {
+ throw new InvalidOperationException("File position must be at 0");
+ }
+
+ // Skip the header
+ if (!reader.FindElement(MatroskaConstants.SegmentContainer))
+ {
+ throw new InvalidOperationException("Expected a segment container");
+ }
+
+ reader.EnterContainer();
+
+ long? tracksPosition = null;
+ long? cuesPosition = null;
+ long? infoPosition = null;
+ // The first element should be a SeekHead otherwise we'll have to search manually
+ if (!reader.FindElement(MatroskaConstants.SeekHead))
+ {
+ throw new InvalidOperationException("Expected a SeekHead");
+ }
+
+ reader.EnterContainer();
+ while (reader.FindElement(MatroskaConstants.Seek))
+ {
+ reader.EnterContainer();
+ reader.ReadNext();
+ var type = (ulong)reader.ReadUIntFromBinary();
+ switch (type)
+ {
+ case MatroskaConstants.Tracks:
+ reader.ReadNext();
+ tracksPosition = (long)reader.ReadUInt();
+ break;
+ case MatroskaConstants.Cues:
+ reader.ReadNext();
+ cuesPosition = (long)reader.ReadUInt();
+ break;
+ case MatroskaConstants.Info:
+ reader.ReadNext();
+ infoPosition = (long)reader.ReadUInt();
+ break;
+ }
+
+ reader.LeaveContainer();
+
+ if (tracksPosition.HasValue && cuesPosition.HasValue && infoPosition.HasValue)
+ {
+ break;
+ }
+ }
+
+ reader.LeaveContainer();
+
+ if (!tracksPosition.HasValue || !cuesPosition.HasValue || !infoPosition.HasValue)
+ {
+ throw new InvalidOperationException("SeekHead is missing or does not contain Info, Tracks and Cues positions");
+ }
+
+ return new SeekHead(infoPosition.Value, tracksPosition.Value, cuesPosition.Value);
+ }
+
+ /// <summary>
+ /// Reads from SegmentContainer to retrieve the Info segment.
+ /// </summary>
+ /// <param name="reader">An instance of <see cref="EbmlReader"/>.</param>
+ /// <returns>Instance of <see cref="Info"/></returns>
+ internal static Info ReadInfo(this EbmlReader reader, long position)
+ {
+ reader.ReadAt(position);
+
+ double? duration = null;
+ reader.EnterContainer();
+ // Mandatory element
+ reader.FindElement(MatroskaConstants.TimestampScale);
+ var timestampScale = reader.ReadUInt();
+
+ if (reader.FindElement(MatroskaConstants.Duration))
+ {
+ duration = reader.ReadFloat();
+ }
+
+ reader.LeaveContainer();
+
+ return new Info((long)timestampScale, duration);
+ }
+
+ /// <summary>
+ /// Enters the Tracks segment and reads all tracks to find the specified type.
+ /// </summary>
+ /// <param name="reader">Instance of <see cref="EbmlReader"/>.</param>
+ /// <param name="tracksPosition">The relative position of the tracks segment.</param>
+ /// <param name="type">The track type identifier.</param>
+ /// <returns>The first track number with the specified type.</returns>
+ /// <exception cref="InvalidOperationException">Stream type is not found.</exception>
+ internal static ulong FindFirstTrackNumberByType(this EbmlReader reader, long tracksPosition, ulong type)
+ {
+ reader.ReadAt(tracksPosition);
+
+ reader.EnterContainer();
+ while (reader.FindElement(MatroskaConstants.TrackEntry))
+ {
+ reader.EnterContainer();
+ // Mandatory element
+ reader.FindElement(MatroskaConstants.TrackNumber);
+ var trackNumber = reader.ReadUInt();
+
+ // Mandatory element
+ reader.FindElement(MatroskaConstants.TrackType);
+ var trackType = reader.ReadUInt();
+
+ reader.LeaveContainer();
+ if (trackType == MatroskaConstants.TrackTypeVideo)
+ {
+ reader.LeaveContainer();
+ return trackNumber;
+ }
+ }
+
+ reader.LeaveContainer();
+
+ throw new InvalidOperationException($"No stream with type {type} found");
+ }
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaConstants.cs b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaConstants.cs
new file mode 100644
index 000000000..d18418d45
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaConstants.cs
@@ -0,0 +1,31 @@
+namespace Jellyfin.MediaEncoding.Keyframes.Matroska
+{
+ /// <summary>
+ /// Constants for the Matroska identifiers.
+ /// </summary>
+ public static class MatroskaConstants
+ {
+ internal const ulong SegmentContainer = 0x18538067;
+
+ internal const ulong SeekHead = 0x114D9B74;
+ internal const ulong Seek = 0x4DBB;
+
+ internal const ulong Info = 0x1549A966;
+ internal const ulong TimestampScale = 0x2AD7B1;
+ internal const ulong Duration = 0x4489;
+
+ internal const ulong Tracks = 0x1654AE6B;
+ internal const ulong TrackEntry = 0xAE;
+ internal const ulong TrackNumber = 0xD7;
+ internal const ulong TrackType = 0x83;
+
+ internal const ulong TrackTypeVideo = 0x1;
+ internal const ulong TrackTypeSubtitle = 0x11;
+
+ internal const ulong Cues = 0x1C53BB6B;
+ internal const ulong CueTime = 0xB3;
+ internal const ulong CuePoint = 0xBB;
+ internal const ulong CueTrackPositions = 0xB7;
+ internal const ulong CuePointTrackNumber = 0xF7;
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaKeyframeExtractor.cs b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaKeyframeExtractor.cs
new file mode 100644
index 000000000..10d017d2a
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/MatroskaKeyframeExtractor.cs
@@ -0,0 +1,76 @@
+using System;
+using System.Collections.Generic;
+using System.IO;
+using Jellyfin.MediaEncoding.Keyframes.Matroska.Extensions;
+using NEbml.Core;
+
+namespace Jellyfin.MediaEncoding.Keyframes.Matroska
+{
+ /// <summary>
+ /// The keyframe extractor for the matroska container.
+ /// </summary>
+ public static class MatroskaKeyframeExtractor
+ {
+ /// <summary>
+ /// Extracts the keyframes in ticks (scaled using the container timestamp scale) from the matroska container.
+ /// </summary>
+ /// <param name="filePath">The file path.</param>
+ /// <returns>An instance of <see cref="KeyframeData"/>.</returns>
+ public static KeyframeData GetKeyframeData(string filePath)
+ {
+ using var stream = File.OpenRead(filePath);
+ using var reader = new EbmlReader(stream);
+
+ var seekHead = reader.ReadSeekHead();
+ var info = reader.ReadInfo(seekHead.InfoPosition);
+ var videoTrackNumber = reader.FindFirstTrackNumberByType(seekHead.TracksPosition, MatroskaConstants.TrackTypeVideo);
+
+ var keyframes = new List<long>();
+ reader.ReadAt(seekHead.CuesPosition);
+ reader.EnterContainer();
+
+ while (reader.FindElement(MatroskaConstants.CuePoint))
+ {
+ reader.EnterContainer();
+ ulong? trackNumber = null;
+ // Mandatory element
+ reader.FindElement(MatroskaConstants.CueTime);
+ var cueTime = reader.ReadUInt();
+
+ // Mandatory element
+ reader.FindElement(MatroskaConstants.CueTrackPositions);
+ reader.EnterContainer();
+ if (reader.FindElement(MatroskaConstants.CuePointTrackNumber))
+ {
+ trackNumber = reader.ReadUInt();
+ }
+
+ reader.LeaveContainer();
+
+ if (trackNumber == videoTrackNumber)
+ {
+ keyframes.Add(ScaleToNanoseconds(cueTime, info.TimestampScale));
+ }
+
+ reader.LeaveContainer();
+ }
+
+ reader.LeaveContainer();
+
+ var result = new KeyframeData(ScaleToNanoseconds(info.Duration ?? 0, info.TimestampScale), keyframes);
+ return result;
+ }
+
+ private static long ScaleToNanoseconds(ulong unscaledValue, long timestampScale)
+ {
+ // TimestampScale is in nanoseconds, scale it to get the value in ticks, 1 tick == 100 ns
+ return (long)unscaledValue * timestampScale / 100;
+ }
+
+ private static long ScaleToNanoseconds(double unscaledValue, long timestampScale)
+ {
+ // TimestampScale is in nanoseconds, scale it to get the value in ticks, 1 tick == 100 ns
+ return Convert.ToInt64(unscaledValue * timestampScale / 100);
+ }
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/Info.cs b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/Info.cs
new file mode 100644
index 000000000..02c6741ec
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/Info.cs
@@ -0,0 +1,29 @@
+namespace Jellyfin.MediaEncoding.Keyframes.Matroska.Models
+{
+ /// <summary>
+ /// The matroska Info segment.
+ /// </summary>
+ internal class Info
+ {
+ /// <summary>
+ /// Initializes a new instance of the <see cref="Info"/> class.
+ /// </summary>
+ /// <param name="timestampScale">The timestamp scale in nanoseconds.</param>
+ /// <param name="duration">The duration of the entire file.</param>
+ public Info(long timestampScale, double? duration)
+ {
+ TimestampScale = timestampScale;
+ Duration = duration;
+ }
+
+ /// <summary>
+ /// Gets the timestamp scale in nanoseconds.
+ /// </summary>
+ public long TimestampScale { get; }
+
+ /// <summary>
+ /// Gets the total duration of the file.
+ /// </summary>
+ public double? Duration { get; }
+ }
+}
diff --git a/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/SeekHead.cs b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/SeekHead.cs
new file mode 100644
index 000000000..d9e346c03
--- /dev/null
+++ b/src/Jellyfin.MediaEncoding.Keyframes/Matroska/Models/SeekHead.cs
@@ -0,0 +1,36 @@
+namespace Jellyfin.MediaEncoding.Keyframes.Matroska.Models
+{
+ /// <summary>
+ /// The matroska SeekHead segment. All positions are relative to the Segment container.
+ /// </summary>
+ internal class SeekHead
+ {
+ /// <summary>
+ /// Initializes a new instance of the <see cref="SeekHead"/> class.
+ /// </summary>
+ /// <param name="infoPosition">The relative file position of the info segment.</param>
+ /// <param name="tracksPosition">The relative file position of the tracks segment.</param>
+ /// <param name="cuesPosition">The relative file position of the cues segment.</param>
+ public SeekHead(long infoPosition, long tracksPosition, long cuesPosition)
+ {
+ InfoPosition = infoPosition;
+ TracksPosition = tracksPosition;
+ CuesPosition = cuesPosition;
+ }
+
+ /// <summary>
+ /// Gets relative file position of the info segment.
+ /// </summary>
+ public long InfoPosition { get; }
+
+ /// <summary>
+ /// Gets the relative file position of the tracks segment.
+ /// </summary>
+ public long TracksPosition { get; }
+
+ /// <summary>
+ /// Gets the relative file position of the cues segment.
+ /// </summary>
+ public long CuesPosition { get; }
+ }
+}