Add AV1 color space parsing in MP4 atom parser

RELEASENOTES.md

@@ -24,6 +24,10 @@
         video track are available. The default value is `false` which means
         selecting a video track is prioritized.
 *   Extractors:
+    *   Add additional AV1C parsing to MP4 extractor to retrieve
+        `ColorInfo.colorSpace`, `ColorInfo.colorTransfer`, and
+        `ColorInfo.colorRange` values
+        ([#692](https://github.com/androidx/media/pull/692)).
 *   Audio:
 *   Video:
     *   Add workaround for a device issue on Galaxy Tab S7 FE, Chromecast with

libraries/extractor/src/main/java/androidx/media3/extractor/mp4/AtomParsers.java

@@ -32,6 +32,7 @@
 import androidx.media3.common.util.CodecSpecificDataUtil;
 import androidx.media3.common.util.Log;
 import androidx.media3.common.util.NullableType;
+import androidx.media3.common.util.ParsableBitArray;
 import androidx.media3.common.util.ParsableByteArray;
 import androidx.media3.common.util.Util;
 import androidx.media3.container.Mp4LocationData;
@@ -1221,22 +1222,15 @@ private static void parseVideoSampleEntry(
         colorTransfer =
             ColorInfo.isoTransferCharacteristicsToColorTransfer(transferCharacteristics);
       } else if (childAtomType == Atom.TYPE_av1C) {
-        ExtractorUtil.checkContainerInput(mimeType == null, /* message= */ null);
         mimeType = MimeTypes.VIDEO_AV1;
         parent.setPosition(childStartPosition + Atom.HEADER_SIZE);
-        parent.skipBytes(1);
-        int byte2 = parent.readUnsignedByte();
-        int seqProfile = byte2 >> 5;
-        int byte3 = parent.readUnsignedByte();
-        boolean highBitdepth = ((byte3 >> 6) & 0b1) != 0;
-        // From https://aomediacodec.github.io/av1-spec/av1-spec.pdf#page=44
-        if (seqProfile == 2 && highBitdepth) {
-          boolean twelveBit = ((byte3 >> 5) & 0b1) != 0;
-          bitdepthLuma = twelveBit ? 12 : 10;
-        } else if (seqProfile <= 2) {
-          bitdepthLuma = highBitdepth ? 10 : 8;
-        }
-        bitdepthChroma = bitdepthLuma;
+        ColorInfo colorInfo = parseAv1c(parent);
+
+        bitdepthLuma = colorInfo.lumaBitdepth;
+        bitdepthChroma = colorInfo.chromaBitdepth;
+        colorSpace = colorInfo.colorSpace;
+        colorRange = colorInfo.colorRange;
+        colorTransfer = colorInfo.colorTransfer;
       } else if (childAtomType == Atom.TYPE_clli) {
         if (hdrStaticInfo == null) {
           hdrStaticInfo = allocateHdrStaticInfo();
@@ -1317,8 +1311,8 @@ private static void parseVideoSampleEntry(
         // established by the bitstream. The absence of color descriptors ('colorSpace' and
         // 'colorTransfer') does not necessarily mean that 'colorRange' has default values, hence it
         // is not being verified here.
-        // If 'Atom.TYPE_avcC', 'Atom.TYPE_hvcC' or 'Atom.TYPE_vpcC' is available, they will take
-        // precedence and overwrite any existing values.
+        // If 'Atom.TYPE_avcC', 'Atom.TYPE_hvcC', 'Atom.TYPE_vpcC' or 'Atom.TYPE_av1c' is available,
+        // they will take precedence and overwrite any existing values.
         if (colorSpace == Format.NO_VALUE && colorTransfer == Format.NO_VALUE) {
           int colorType = parent.readInt();
           if (colorType == TYPE_nclx || colorType == TYPE_nclc) {
@@ -1385,6 +1379,138 @@ private static void parseVideoSampleEntry(
     out.format = formatBuilder.build();
   }
 
+  /**
+   * Parses the av1C configuration record and OBU sequence header and returns a {@link ColorInfo}
+   * from their data.
+   *
+   * <p>See av1C configuration record syntax in this <a
+   * href="https://aomediacodec.github.io/av1-isobmff/#av1codecconfigurationbox-syntax">spec</a>.
+   *
+   * <p>See av1C OBU syntax in this <a
+   * href="https://aomediacodec.github.io/av1-spec/av1-spec.pdf">spec</a>.
+   *
+   * <p>The sections referenced in the method are from these specs.
+   *
+   * @param data The av1C atom data.
+   * @return {@link ColorInfo} parsed from the av1C data.
+   */
+  private static ColorInfo parseAv1c(ParsableByteArray data) {
+    ColorInfo.Builder colorInfo = new ColorInfo.Builder();
+    ParsableBitArray bitArray = new ParsableBitArray(data.getData());
+    bitArray.setPosition(data.getPosition() * 8); // Convert byte to bit position.
+
+    // Parse av1C config record for bitdepth info.
+    // See https://aomediacodec.github.io/av1-isobmff/#av1codecconfigurationbox-syntax.
+    bitArray.skipBytes(1); // marker, version
+    int seqProfile = bitArray.readBits(3); // seq_profile
+    bitArray.skipBits(6); // seq_level_idx_0, seq_tier_0
+    boolean highBitdepth = bitArray.readBit(); // high_bitdepth
+    boolean twelveBit = bitArray.readBit(); // twelve_bit
+    if (seqProfile == 2 && highBitdepth) {
+      colorInfo.setLumaBitdepth(twelveBit ? 12 : 10);
+      colorInfo.setChromaBitdepth(twelveBit ? 12 : 10);
+    } else if (seqProfile <= 2) {
+      colorInfo.setLumaBitdepth(highBitdepth ? 10 : 8);
+      colorInfo.setChromaBitdepth(highBitdepth ? 10 : 8);
+    }
+    // Skip monochrome, chroma_subsampling_x, chroma_subsampling_y, chroma_sample_position,
+    // reserved and initial_presentation_delay.
+    bitArray.skipBits(13);
+
+    // 5.3.1. General OBU syntax
+    bitArray.skipBit(); // obu_forbidden_bit
+    int obuType = bitArray.readBits(4); // obu_type
+    if (obuType != 1) { // obu_type != OBU_SEQUENCE_HEADER
+      Log.i(TAG, "Unsupported obu_type: " + obuType);
+      return colorInfo.build();
+    }
+    if (bitArray.readBit()) { // obu_extension_flag
+      Log.i(TAG, "Unsupported obu_extension_flag");
+      return colorInfo.build();
+    }
+    boolean obuHasSizeField = bitArray.readBit(); // obu_has_size_field
+    bitArray.skipBit(); // obu_reserved_1bit
+    // obu_size is unsigned leb128 and if obu_size <= 127 then it can be simplified as readBits(8).
+    if (obuHasSizeField && bitArray.readBits(8) > 127) { // obu_size
+      Log.i(TAG, "Excessive obu_size");
+      return colorInfo.build();
+    }
+    // 5.5.1. General OBU sequence header syntax
+    int obuSeqHeaderSeqProfile = bitArray.readBits(3); // seq_profile
+    bitArray.skipBit(); // still_picture
+    if (bitArray.readBit()) { // reduced_still_picture_header
+      Log.i(TAG, "Unsupported reduced_still_picture_header");
+      return colorInfo.build();
+    }
+    if (bitArray.readBit()) { // timing_info_present_flag
+      Log.i(TAG, "Unsupported timing_info_present_flag");
+      return colorInfo.build();
+    }
+    if (bitArray.readBit()) { // initial_display_delay_present_flag
+      Log.i(TAG, "Unsupported initial_display_delay_present_flag");
+      return colorInfo.build();
+    }
+    int operatingPointsCountMinus1 = bitArray.readBits(5); // operating_points_cnt_minus_1
+    for (int i = 0; i <= operatingPointsCountMinus1; i++) {
+      bitArray.skipBits(12); // operating_point_idc[i]
+      int seqLevelIdx = bitArray.readBits(5); // seq_level_idx[i]
+      if (seqLevelIdx > 7) {
+        bitArray.skipBit(); // seq_tier[i]
+      }
+    }
+    int frameWidthBitsMinus1 = bitArray.readBits(4); // frame_width_bits_minus_1
+    int frameHeightBitsMinus1 = bitArray.readBits(4); // frame_height_bits_minus_1
+    bitArray.skipBits(frameWidthBitsMinus1 + 1); // max_frame_width_minus_1
+    bitArray.skipBits(frameHeightBitsMinus1 + 1); // max_frame_height_minus_1
+    if (bitArray.readBit()) { // frame_id_numbers_present_flag
+      bitArray.skipBits(7); // delta_frame_id_length_minus_2, additional_frame_id_length_minus_1
+    }
+    bitArray.skipBits(7); // use_128x128_superblock...enable_dual_filter: 7 flags
+    boolean enableOrderHint = bitArray.readBit(); // enable_order_hint
+    if (enableOrderHint) {
+      bitArray.skipBits(2); // enable_jnt_comp, enable_ref_frame_mvs
+    }
+    int seqForceScreenContentTools =
+        bitArray.readBit() // seq_choose_screen_content_tools
+            ? 2 // SELECT_SCREEN_CONTENT_TOOLS
+            : bitArray.readBits(1); // seq_force_screen_content_tools
+    if (seqForceScreenContentTools > 0) {
+      if (!bitArray.readBit()) { // seq_choose_integer_mv
+        bitArray.skipBits(1); // seq_force_integer_mv
+      }
+    }
+    if (enableOrderHint) {
+      bitArray.skipBits(3); // order_hint_bits_minus_1
+    }
+    bitArray.skipBits(3); // enable_superres, enable_cdef, enable_restoration
+    // 5.5.2. OBU Color config syntax
+    boolean colorConfigHighBitdepth = bitArray.readBit(); // high_bitdepth
+    if (obuSeqHeaderSeqProfile == 2 && colorConfigHighBitdepth) {
+      bitArray.skipBit(); // twelve_bit
+    }
+
+    boolean monochrome = (obuSeqHeaderSeqProfile != 1) && bitArray.readBit(); // mono_chrome
+
+    if (bitArray.readBit()) { // color_description_present_flag
+      int colorPrimaries = bitArray.readBits(8); // color_primaries
+      int transferCharacteristics = bitArray.readBits(8); // transfer_characteristics
+      int matrixCoefficients = bitArray.readBits(8); // matrix_coefficients
+      int colorRange =
+          (!monochrome
+                  && colorPrimaries == 1 // CP_BT_709
+                  && transferCharacteristics == 13 // TC_SRGB
+                  && matrixCoefficients == 0) // MC_IDENTITY
+              ? 1
+              : bitArray.readBits(1); // color_range;
+      colorInfo
+          .setColorSpace(ColorInfo.isoColorPrimariesToColorSpace(colorPrimaries))
+          .setColorRange((colorRange == 1) ? C.COLOR_RANGE_FULL : C.COLOR_RANGE_LIMITED)
+          .setColorTransfer(
+              ColorInfo.isoTransferCharacteristicsToColorTransfer(transferCharacteristics));
+    }
+    return colorInfo.build();
+  }
+
   private static ByteBuffer allocateHdrStaticInfo() {
     // For HDR static info, Android decoders expect a 25-byte array. The first byte is zero to
     // represent Static Metadata Type 1, as per CTA-861-G:2017, Table 44. The following 24 bytes

libraries/extractor/src/test/java/androidx/media3/extractor/mp4/Mp4ExtractorTest.java

@@ -134,4 +134,10 @@ public void mp4SampleWithOriginalQuicktimeSpecification() throws Exception {
         "media/mp4/sample_with_original_quicktime_specification.mov",
         simulationConfig);
   }
+
+  @Test
+  public void mp4SampleWithAv1c() throws Exception {
+    ExtractorAsserts.assertBehavior(
+        Mp4Extractor::new, "media/mp4/sample_with_av1c.mp4", simulationConfig);
+  }
 }

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_av1c.mp4.0.dump

@@ -0,0 +1,146 @@
+seekMap:
+  isSeekable = true
+  duration = 1000000
+  getPosition(0) = [[timeUs=0, position=48]]
+  getPosition(1) = [[timeUs=0, position=48]]
+  getPosition(500000) = [[timeUs=0, position=48]]
+  getPosition(1000000) = [[timeUs=0, position=48]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 942
+  sample count = 30
+  format 0:
+    id = 1
+    sampleMimeType = video/av01
+    maxInputSize = 188
+    width = 720
+    height = 1280
+    frameRate = 30.0
+    colorInfo:
+      colorSpace = 6
+      colorRange = 2
+      colorTransfer = 7
+      lumaBitdepth = 10
+      chromaBitdepth = 10
+    metadata = entries=[TSSE: description=null: values=[Lavf60.3.100], Mp4Timestamp: creation time=0, modification time=0, timescale=1000]
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 84, hash 9C46A819
+  sample 1:
+    time = 33333
+    flags = 0
+    data = length 158, hash 43A1B544
+  sample 2:
+    time = 66666
+    flags = 0
+    data = length 3, hash D600
+  sample 3:
+    time = 100000
+    flags = 0
+    data = length 28, hash 27890E81
+  sample 4:
+    time = 133333
+    flags = 0
+    data = length 3, hash D5F0
+  sample 5:
+    time = 166666
+    flags = 0
+    data = length 55, hash 9FC5012E
+  sample 6:
+    time = 200000
+    flags = 0
+    data = length 3, hash D600
+  sample 7:
+    time = 233333
+    flags = 0
+    data = length 27, hash 70CFAC05
+  sample 8:
+    time = 266666
+    flags = 0
+    data = length 3, hash D5D0
+  sample 9:
+    time = 300000
+    flags = 0
+    data = length 82, hash 944218D6
+  sample 10:
+    time = 333333
+    flags = 0
+    data = length 3, hash D600
+  sample 11:
+    time = 366666
+    flags = 0
+    data = length 27, hash BA4D4A06
+  sample 12:
+    time = 400000
+    flags = 0
+    data = length 3, hash D5F0
+  sample 13:
+    time = 433333
+    flags = 0
+    data = length 54, hash A98584CA
+  sample 14:
+    time = 466666
+    flags = 0
+    data = length 3, hash D600
+  sample 15:
+    time = 500000
+    flags = 0
+    data = length 27, hash 45D733B8
+  sample 16:
+    time = 533333
+    flags = 0
+    data = length 3, hash D5A0
+  sample 17:
+    time = 566666
+    flags = 0
+    data = length 112, hash B80B26FD
+  sample 18:
+    time = 600000
+    flags = 0
+    data = length 3, hash D5F0
+  sample 19:
+    time = 633333
+    flags = 0
+    data = length 27, hash 37DD29D9
+  sample 20:
+    time = 666666
+    flags = 0
+    data = length 3, hash D5E0
+  sample 21:
+    time = 700000
+    flags = 0
+    data = length 54, hash 1C15581C
+  sample 22:
+    time = 733333
+    flags = 0
+    data = length 3, hash D5F0
+  sample 23:
+    time = 766666
+    flags = 0
+    data = length 27, hash 49EC3531
+  sample 24:
+    time = 800000
+    flags = 0
+    data = length 3, hash D5B0
+  sample 25:
+    time = 833333
+    flags = 0
+    data = length 84, hash 2025C9F5
+  sample 26:
+    time = 866666
+    flags = 0
+    data = length 3, hash D5D0
+  sample 27:
+    time = 900000
+    flags = 0
+    data = length 27, hash B927669C
+  sample 28:
+    time = 933333
+    flags = 0
+    data = length 3, hash D5C0
+  sample 29:
+    time = 966666
+    flags = 536870912
+    data = length 27, hash 706C58AD
+tracksEnded = true