Merge pull request #35 from ittiam-systems:rtp-mpeg4

PiperOrigin-RevId: 438000682
androidx · Apr 6, 2022 · f48babb · f48babb
2 parents 8089092 + ef9393a
commit f48babb
Show file tree

Hide file tree

Showing 6 changed files with 298 additions and 6 deletions.
diff --git a/RELEASENOTES.md b/RELEASENOTES.md
@@ -36,6 +36,8 @@
         dependency from the UI module to the ExoPlayer module. This is a
         breaking change.
 *   RTSP:
+    *   Add RTP reader for MPEG4
+        ([#35](https://github.com/androidx/media/pull/35))
     *   Add RTP reader for HEVC
         ([#36](https://github.com/androidx/media/pull/36)).
     *   Add RTP reader for AMR. Currently only mono-channel, non-interleaved

diff --git a/libraries/common/src/main/java/androidx/media3/common/util/CodecSpecificDataUtil.java b/libraries/common/src/main/java/androidx/media3/common/util/CodecSpecificDataUtil.java
@@ -15,6 +15,8 @@
  */
 package androidx.media3.common.util;
 
+import static androidx.media3.common.util.Assertions.checkArgument;
+
 import android.util.Pair;
 import androidx.annotation.Nullable;
 import androidx.media3.common.C;
@@ -31,6 +33,12 @@ public final class CodecSpecificDataUtil {
   private static final String[] HEVC_GENERAL_PROFILE_SPACE_STRINGS =
       new String[] {"", "A", "B", "C"};
 
+  // MP4V-ES
+  private static final int VISUAL_OBJECT_LAYER = 1;
+  private static final int VISUAL_OBJECT_LAYER_START = 0x20;
+  private static final int EXTENDED_PAR = 0x0F;
+  private static final int RECTANGULAR = 0x00;
+
   /**
    * Parses an ALAC AudioSpecificConfig (i.e. an <a
    * href="https://github.com/macosforge/alac/blob/master/ALACMagicCookieDescription.txt">ALACSpecificConfig</a>).
@@ -72,6 +80,87 @@ public static boolean parseCea708InitializationData(List<byte[]> initializationD
         && initializationData.get(0)[0] == 1;
   }
 
+  /**
+   * Parses an MPEG-4 Visual configuration information, as defined in ISO/IEC14496-2.
+   *
+   * @param videoSpecificConfig A byte array containing the MPEG-4 Visual configuration information
+   *     to parse.
+   * @return A pair of the video's width and height.
+   */
+  public static Pair<Integer, Integer> getVideoResolutionFromMpeg4VideoConfig(
+      byte[] videoSpecificConfig) {
+    int offset = 0;
+    boolean foundVOL = false;
+    ParsableByteArray scratchBytes = new ParsableByteArray(videoSpecificConfig);
+    while (offset + 3 < videoSpecificConfig.length) {
+      if (scratchBytes.readUnsignedInt24() != VISUAL_OBJECT_LAYER
+          || (videoSpecificConfig[offset + 3] & 0xF0) != VISUAL_OBJECT_LAYER_START) {
+        scratchBytes.setPosition(scratchBytes.getPosition() - 2);
+        offset++;
+        continue;
+      }
+      foundVOL = true;
+      break;
+    }
+
+    checkArgument(foundVOL, "Invalid input: VOL not found.");
+
+    ParsableBitArray scratchBits = new ParsableBitArray(videoSpecificConfig);
+    // Skip the start codecs from the bitstream
+    scratchBits.skipBits((offset + 4) * 8);
+    scratchBits.skipBits(1); // random_accessible_vol
+    scratchBits.skipBits(8); // video_object_type_indication
+
+    if (scratchBits.readBit()) { // object_layer_identifier
+      scratchBits.skipBits(4); // video_object_layer_verid
+      scratchBits.skipBits(3); // video_object_layer_priority
+    }
+
+    int aspectRatioInfo = scratchBits.readBits(4);
+    if (aspectRatioInfo == EXTENDED_PAR) {
+      scratchBits.skipBits(8); // par_width
+      scratchBits.skipBits(8); // par_height
+    }
+
+    if (scratchBits.readBit()) { // vol_control_parameters
+      scratchBits.skipBits(2); // chroma_format
+      scratchBits.skipBits(1); // low_delay
+      if (scratchBits.readBit()) { // vbv_parameters
+        scratchBits.skipBits(79);
+      }
+    }
+
+    int videoObjectLayerShape = scratchBits.readBits(2);
+    checkArgument(
+        videoObjectLayerShape == RECTANGULAR,
+        "Only supports rectangular video object layer shape.");
+
+    checkArgument(scratchBits.readBit()); // marker_bit
+    int vopTimeIncrementResolution = scratchBits.readBits(16);
+    checkArgument(scratchBits.readBit()); // marker_bit
+
+    if (scratchBits.readBit()) { // fixed_vop_rate
+      checkArgument(vopTimeIncrementResolution > 0);
+      vopTimeIncrementResolution--;
+      int numBitsToSkip = 0;
+      while (vopTimeIncrementResolution > 0) {
+        numBitsToSkip++;
+        vopTimeIncrementResolution >>= 1;
+      }
+      scratchBits.skipBits(numBitsToSkip); // fixed_vop_time_increment
+    }
+
+    checkArgument(scratchBits.readBit()); // marker_bit
+    int videoObjectLayerWidth = scratchBits.readBits(13);
+    checkArgument(scratchBits.readBit()); // marker_bit
+    int videoObjectLayerHeight = scratchBits.readBits(13);
+    checkArgument(scratchBits.readBit()); // marker_bit
+
+    scratchBits.skipBits(1); // interlaced
+
+    return Pair.create(videoObjectLayerWidth, videoObjectLayerHeight);
+  }
+
   /**
    * Builds an RFC 6381 AVC codec string using the provided parameters.
    *

diff --git a/libraries/exoplayer_rtsp/src/main/java/androidx/media3/exoplayer/rtsp/RtpPayloadFormat.java b/libraries/exoplayer_rtsp/src/main/java/androidx/media3/exoplayer/rtsp/RtpPayloadFormat.java
@@ -43,6 +43,7 @@ public final class RtpPayloadFormat {
   private static final String RTP_MEDIA_AMR = "AMR";
   private static final String RTP_MEDIA_AMR_WB = "AMR-WB";
   private static final String RTP_MEDIA_MPEG4_GENERIC = "MPEG4-GENERIC";
+  private static final String RTP_MEDIA_MPEG4_VIDEO = "MP4V-ES";
   private static final String RTP_MEDIA_H264 = "H264";
   private static final String RTP_MEDIA_H265 = "H265";
   private static final String RTP_MEDIA_PCM_L8 = "L8";
@@ -59,6 +60,7 @@ public static boolean isFormatSupported(MediaDescription mediaDescription) {
       case RTP_MEDIA_AMR_WB:
       case RTP_MEDIA_H264:
       case RTP_MEDIA_H265:
+      case RTP_MEDIA_MPEG4_VIDEO:
       case RTP_MEDIA_MPEG4_GENERIC:
       case RTP_MEDIA_PCM_L8:
       case RTP_MEDIA_PCM_L16:
@@ -86,10 +88,6 @@ public static String getMimeTypeFromRtpMediaType(String mediaType) {
         return MimeTypes.AUDIO_AMR_NB;
       case RTP_MEDIA_AMR_WB:
         return MimeTypes.AUDIO_AMR_WB;
-      case RTP_MEDIA_H264:
-        return MimeTypes.VIDEO_H264;
-      case RTP_MEDIA_H265:
-        return MimeTypes.VIDEO_H265;
       case RTP_MEDIA_MPEG4_GENERIC:
         return MimeTypes.AUDIO_AAC;
       case RTP_MEDIA_PCM_L8:
@@ -99,6 +97,12 @@ public static String getMimeTypeFromRtpMediaType(String mediaType) {
         return MimeTypes.AUDIO_ALAW;
       case RTP_MEDIA_PCMU:
         return MimeTypes.AUDIO_MLAW;
+      case RTP_MEDIA_H264:
+        return MimeTypes.VIDEO_H264;
+      case RTP_MEDIA_H265:
+        return MimeTypes.VIDEO_H265;
+      case RTP_MEDIA_MPEG4_VIDEO:
+        return MimeTypes.VIDEO_MP4V;
       case RTP_MEDIA_VP8:
         return MimeTypes.VIDEO_VP8;
       default:

diff --git a/libraries/exoplayer_rtsp/src/main/java/androidx/media3/exoplayer/rtsp/RtspMediaTrack.java b/libraries/exoplayer_rtsp/src/main/java/androidx/media3/exoplayer/rtsp/RtspMediaTrack.java
@@ -25,6 +25,7 @@
 
 import android.net.Uri;
 import android.util.Base64;
+import android.util.Pair;
 import androidx.annotation.Nullable;
 import androidx.annotation.VisibleForTesting;
 import androidx.media3.common.C;
@@ -44,19 +45,42 @@
   // Format specific parameter names.
   private static final String PARAMETER_PROFILE_LEVEL_ID = "profile-level-id";
   private static final String PARAMETER_SPROP_PARAMS = "sprop-parameter-sets";
+
+  private static final String PARAMETER_AMR_OCTET_ALIGN = "octet-align";
+  private static final String PARAMETER_AMR_INTERLEAVING = "interleaving";
   private static final String PARAMETER_H265_SPROP_SPS = "sprop-sps";
   private static final String PARAMETER_H265_SPROP_PPS = "sprop-pps";
   private static final String PARAMETER_H265_SPROP_VPS = "sprop-vps";
   private static final String PARAMETER_H265_SPROP_MAX_DON_DIFF = "sprop-max-don-diff";
-  private static final String PARAMETER_AMR_OCTET_ALIGN = "octet-align";
-  private static final String PARAMETER_AMR_INTERLEAVING = "interleaving";
+  private static final String PARAMETER_MP4V_CONFIG = "config";
 
   /** Prefix for the RFC6381 codecs string for AAC formats. */
   private static final String AAC_CODECS_PREFIX = "mp4a.40.";
   /** Prefix for the RFC6381 codecs string for AVC formats. */
   private static final String H264_CODECS_PREFIX = "avc1.";
+  /** Prefix for the RFC6416 codecs string for MPEG4V-ES formats. */
+  private static final String MPEG4_CODECS_PREFIX = "mp4v.";
 
   private static final String GENERIC_CONTROL_ATTR = "*";
+  /**
+   * Default height for MP4V.
+   *
+   * <p>RFC6416 does not mandate codec specific data (like width and height) in the fmtp attribute.
+   * These values are taken from <a
+   * href=https://cs.android.com/android/platform/superproject/+/master:frameworks/av/media/codec2/components/mpeg4_h263/C2SoftMpeg4Dec.cpp;l=130
+   * >Android's software MP4V decoder</a>.
+   */
+  private static final int DEFAULT_MP4V_WIDTH = 352;
+
+  /**
+   * Default height for MP4V.
+   *
+   * <p>RFC6416 does not mandate codec specific data (like width and height) in the fmtp attribute.
+   * These values are taken from <a
+   * href=https://cs.android.com/android/platform/superproject/+/master:frameworks/av/media/codec2/components/mpeg4_h263/C2SoftMpeg4Dec.cpp;l=130
+   * >Android's software MP4V decoder</a>.
+   */
+  private static final int DEFAULT_MP4V_HEIGHT = 288;
 
   /**
    * Default width for VP8.
@@ -156,6 +180,10 @@ public int hashCode() {
             !fmtpParameters.containsKey(PARAMETER_AMR_INTERLEAVING),
             "Interleaving mode is not currently supported.");
         break;
+      case MimeTypes.VIDEO_MP4V:
+        checkArgument(!fmtpParameters.isEmpty());
+        processMPEG4FmtpAttribute(formatBuilder, fmtpParameters);
+        break;
       case MimeTypes.VIDEO_H264:
         checkArgument(!fmtpParameters.isEmpty());
         processH264FmtpAttribute(formatBuilder, fmtpParameters);
@@ -214,6 +242,23 @@ private static void processAacFmtpAttribute(
             AacUtil.buildAacLcAudioSpecificConfig(sampleRate, channelCount)));
   }
 
+  private static void processMPEG4FmtpAttribute(
+      Format.Builder formatBuilder, ImmutableMap<String, String> fmtpAttributes) {
+    @Nullable String configInput = fmtpAttributes.get(PARAMETER_MP4V_CONFIG);
+    if (configInput != null) {
+      byte[] configBuffer = Util.getBytesFromHexString(configInput);
+      formatBuilder.setInitializationData(ImmutableList.of(configBuffer));
+      Pair<Integer, Integer> resolution =
+          CodecSpecificDataUtil.getVideoResolutionFromMpeg4VideoConfig(configBuffer);
+      formatBuilder.setWidth(resolution.first).setHeight(resolution.second);
+    } else {
+      // set the default width and height
+      formatBuilder.setWidth(DEFAULT_MP4V_WIDTH).setHeight(DEFAULT_MP4V_HEIGHT);
+    }
+    @Nullable String profileLevel = fmtpAttributes.get(PARAMETER_PROFILE_LEVEL_ID);
+    formatBuilder.setCodecs(MPEG4_CODECS_PREFIX + (profileLevel == null ? "1" : profileLevel));
+  }
+
   /** Returns H264/H265 initialization data from the RTP parameter set. */
   private static byte[] getInitializationDataFromParameterSet(String parameterSet) {
     byte[] decodedParameterNalData = Base64.decode(parameterSet, Base64.DEFAULT);

diff --git a/...p/src/main/java/androidx/media3/exoplayer/rtsp/reader/DefaultRtpPayloadReaderFactory.java b/...p/src/main/java/androidx/media3/exoplayer/rtsp/reader/DefaultRtpPayloadReaderFactory.java
@@ -47,6 +47,8 @@ public RtpPayloadReader createPayloadReader(RtpPayloadFormat payloadFormat) {
         return new RtpH264Reader(payloadFormat);
       case MimeTypes.VIDEO_H265:
         return new RtpH265Reader(payloadFormat);
+      case MimeTypes.VIDEO_MP4V:
+        return new RtpMpeg4Reader(payloadFormat);
       case MimeTypes.VIDEO_VP8:
         return new RtpVp8Reader(payloadFormat);
       default: