feat(CEA): Parse CEA from TS with H.265 (#5610)

shaka-project · Sep 7, 2023 · 54eaf63 · 54eaf63
1 parent fc93292
commit 54eaf63
Show file tree

Hide file tree

Showing 3 changed files with 52 additions and 12 deletions.
diff --git a/lib/cea/ts_cea_parser.js b/lib/cea/ts_cea_parser.js
@@ -49,10 +49,23 @@ shaka.cea.TsCeaParser = class {
       return captionPackets;
     }
     const tsParser = new shaka.util.TsParser().parse(uint8ArrayData);
+    const codecs = tsParser.getCodecs();
     const videoNalus = tsParser.getVideoNalus();
+    const validNaluTypes = [];
+    switch (codecs.video) {
+      case 'avc':
+        validNaluTypes.push(CeaUtils.H264_NALU_TYPE_SEI);
+        break;
+      case 'hvc':
+        validNaluTypes.push(CeaUtils.H265_PREFIX_NALU_TYPE_SEI);
+        validNaluTypes.push(CeaUtils.H265_SUFFIX_NALU_TYPE_SEI);
+        break;
+    }
+    if (!validNaluTypes.length) {
+      return captionPackets;
+    }
     for (const nalu of videoNalus) {
-      if (nalu.type == CeaUtils.H264_NALU_TYPE_SEI &&
-          nalu.time != null) {
+      if (validNaluTypes.includes(nalu.type) && nalu.time != null) {
         for (const packet of this.seiProcessor_.process(nalu.data)) {
           captionPackets.push({
             packet: packet,

diff --git a/lib/transmuxer/ts_transmuxer.js b/lib/transmuxer/ts_transmuxer.js
@@ -632,7 +632,7 @@ shaka.transmuxer.TsTransmuxer = class {
       if (i + 1 < videoData.length) {
         nextPes = videoData[i + 1];
       }
-      const dataNalus = tsParser.parseAvcNalus(pes, nextPes);
+      const dataNalus = tsParser.parseNalus(pes, nextPes);
       nalus = nalus.concat(dataNalus);
       const frame = H264.parseFrame(dataNalus);
       if (!frame) {

diff --git a/lib/util/ts_parser.js b/lib/util/ts_parser.js
@@ -416,11 +416,34 @@ shaka.util.TsParser = class {
    * @export
    */
   parseAvcNalus(pes, nextPes) {
+    shaka.Deprecate.deprecateFeature(5,
+        'TsParser',
+        'Please use parseNalus function instead.');
+    return this.parseNalus(pes, nextPes);
+  }
+
+  /**
+   * Parse AVC and HVC Nalus
+   *
+   * The code is based on hls.js
+   * Credit to https://github.com/video-dev/hls.js/blob/master/src/demux/tsdemuxer.ts
+   *
+   * @param {shaka.extern.MPEG_PES} pes
+   * @param {?shaka.extern.MPEG_PES=} nextPes
+   * @return {!Array.<shaka.extern.VideoNalu>}
+   * @export
+   */
+  parseNalus(pes, nextPes) {
     const timescale = shaka.util.TsParser.Timescale;
     const time = pes.pts ? pes.pts / timescale : null;
     let data = pes.data;
     let len = data.byteLength;
 
+    let naluHeaderSize = 1;
+    if (this.videoCodec_ == 'hvc') {
+      naluHeaderSize = 2;
+    }
+
     // A NALU does not contain is its size.
     // The Annex B specification solves this by requiring ‘Start Codes’ to
     // precede each NALU. A start code is 2 or 3 0x00 bytes followed with a
@@ -459,8 +482,8 @@ shaka.util.TsParser = class {
       } else if (numZeros >= 2 && value == 1) {
         // We just read a start code.  Consume the NALU we passed, if any.
         if (lastNaluStart >= 0) {
-          // Because the start position includes the type, skip the first byte.
-          const firstByteToKeep = lastNaluStart + 1;
+          // Because the start position includes the header size.
+          const firstByteToKeep = lastNaluStart + naluHeaderSize;
 
           // Compute the last byte to keep.  The start code is at most 3 zeros.
           // Any earlier zeros are not part of the start code.
@@ -488,7 +511,11 @@ shaka.util.TsParser = class {
         // Advance and read the type of the next NALU.
         i++;
         lastNaluStart = i;
-        lastNaluType = data[i] & 0x1f;
+        if (this.videoCodec_ == 'hvc') {
+          lastNaluType = (data[i] >> 1) & 0x3f;
+        } else {
+          lastNaluType = data[i] & 0x1f;
+        }
         numZeros = 0;
       } else {
         numZeros = 0;
@@ -505,8 +532,8 @@ shaka.util.TsParser = class {
         } else {
           tryToFinishLastNalu = true;
           // The rest of the buffer was a NALU.
-          // Because the start position includes the type, skip the first byte.
-          const firstByteToKeep = lastNaluStart + 1;
+          // Because the start position includes the header size.
+          const firstByteToKeep = lastNaluStart + naluHeaderSize;
           infoOfLastNalu = {
             data: data.subarray(firstByteToKeep, len),
             fullData: data.subarray(lastNaluStart, len),
@@ -639,17 +666,14 @@ shaka.util.TsParser = class {
    */
   getVideoNalus() {
     const nalus = [];
-    if (this.videoCodec_ != 'avc') {
-      return nalus;
-    }
     const videoData = this.getVideoData();
     for (let i = 0; i < videoData.length; i++) {
       const pes = videoData[i];
       let nextPes;
       if (i + 1 < videoData.length) {
         nextPes = videoData[i + 1];
       }
-      nalus.push(...this.parseAvcNalus(pes, nextPes));
+      nalus.push(...this.parseNalus(pes, nextPes));
     }
     return nalus;
   }
@@ -684,6 +708,9 @@ shaka.util.TsParser = class {
       width: null,
       codec: null,
     };
+    if (this.videoCodec_ != 'avc') {
+      return videoInfo;
+    }
     const videoNalus = this.getVideoNalus();
     if (!videoNalus.length) {
       return videoInfo;