vpxenc: Add support for pixel aspect ratio.

WebM files will adjust the display width and height according to the input pixel aspect ratio. The default pixel aspect ratio is 1:1. BUG=https://code.google.com/p/webm/issues/detail?id=1005 Change-Id: I23e0a601b7259fa9513cb86110c41b8437769808
ittiamvpx · Jun 5, 2015 · 09acd26 · 09acd26
1 parent 8710cce
commit 09acd26
Show file tree

Hide file tree

Showing 8 changed files with 62 additions and 7 deletions.
diff --git a/test/test-data.mk b/test/test-data.mk
@@ -12,6 +12,7 @@ LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_420.y4m
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_422.y4m
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_444.y4m
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_12_440.yuv
+LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_420_a10-1.y4m
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_420.y4m
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_422.y4m
 LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_444.y4m

diff --git a/test/test-data.sha1 b/test/test-data.sha1
@@ -22,6 +22,7 @@ c934da6fb8cc54ee2a8c17c54cf6076dac37ead0 *park_joy_90p_10_440.yuv
 c92825f1ea25c5c37855083a69faac6ac4641a9e *park_joy_90p_12_422.y4m
 b592189b885b6cc85db55cc98512a197d73d3b34 *park_joy_90p_12_444.y4m
 82c1bfcca368c2f22bad7d693d690d5499ecdd11 *park_joy_90p_12_440.yuv
+b9e1e90aece2be6e2c90d89e6ab2372d5f8c792d *park_joy_90p_8_420_a10-1.y4m
 4e0eb61e76f0684188d9bc9f3ce61f6b6b77bb2c *park_joy_90p_8_420.y4m
 7a193ff7dfeb96ba5f82b2afd7afa9e1fe83d947 *park_joy_90p_8_422.y4m
 bdb7856e6bc93599bdda05c2e773a9f22b6c6d03 *park_joy_90p_8_444.y4m

diff --git a/test/tools_common.sh b/test/tools_common.sh
@@ -408,6 +408,8 @@ YUV_RAW_INPUT="${LIBVPX_TEST_DATA_PATH}/hantro_collage_w352h288.yuv"
 YUV_RAW_INPUT_WIDTH=352
 YUV_RAW_INPUT_HEIGHT=288
 
+Y4M_NOSQ_PAR_INPUT="${LIBVPX_TEST_DATA_PATH}/park_joy_90p_8_420_a10-1.y4m"
+
 # Setup a trap function to clean up after tests complete.
 trap cleanup EXIT
 
@@ -429,6 +431,7 @@ vlog "$(basename "${0%.*}") test configuration:
   VPX_TEST_VERBOSE_OUTPUT=${VPX_TEST_VERBOSE_OUTPUT}
   YUV_RAW_INPUT=${YUV_RAW_INPUT}
   YUV_RAW_INPUT_WIDTH=${YUV_RAW_INPUT_WIDTH}
-  YUV_RAW_INPUT_HEIGHT=${YUV_RAW_INPUT_HEIGHT}"
+  YUV_RAW_INPUT_HEIGHT=${YUV_RAW_INPUT_HEIGHT}
+  Y4M_NOSQ_PAR_INPUT=${Y4M_NOSQ_PAR_INPUT}"
 
 fi  # End $VPX_TEST_TOOLS_COMMON_SH pseudo include guard.
diff --git a/test/vpxenc.sh b/test/vpxenc.sh
@@ -23,6 +23,13 @@ vpxenc_verify_environment() {
     elog "The file ${YUV_RAW_INPUT##*/} must exist in LIBVPX_TEST_DATA_PATH."
     return 1
   fi
+  if [ "$(vpxenc_can_encode_vp9)" = "yes" ]; then
+    if [ ! -e "${Y4M_NOSQ_PAR_INPUT}" ]; then
+      elog "The file ${Y4M_NOSQ_PAR_INPUT##*/} must exist in"
+      elog "LIBVPX_TEST_DATA_PATH."
+      return 1
+    fi
+  fi
   if [ -z "$(vpx_tool_path vpxenc)" ]; then
     elog "vpxenc not found. It must exist in LIBVPX_BIN_PATH or its parent."
     return 1
@@ -49,6 +56,10 @@ yuv_input_hantro_collage() {
        --height="${YUV_RAW_INPUT_HEIGHT}""
 }
 
+y4m_input_non_square_par() {
+  echo ""${Y4M_NOSQ_PAR_INPUT}""
+}
+
 # Echo default vpxenc real time encoding params. $1 is the codec, which defaults
 # to vp8 if unspecified.
 vpxenc_rt_params() {
@@ -320,6 +331,23 @@ vpxenc_vp9_webm_lag10_frames20() {
   fi
 }
 
+# TODO(fgalligan): Test that DisplayWidth is different than video width.
+vpxenc_vp9_webm_non_square_par() {
+  if [ "$(vpxenc_can_encode_vp9)" = "yes" ] && \
+     [ "$(webm_io_available)" = "yes" ]; then
+    local readonly output="${VPX_TEST_OUTPUT_DIR}/vp9_non_square_par.webm"
+    vpxenc $(y4m_input_non_square_par) \
+      --codec=vp9 \
+      --limit="${TEST_FRAMES}" \
+      --output="${output}"
+
+    if [ ! -e "${output}" ]; then
+      elog "Output file does not exist."
+      return 1
+    fi
+  fi
+}
+
 vpxenc_tests="vpxenc_vp8_ivf
               vpxenc_vp8_webm
               vpxenc_vp8_webm_rt
@@ -332,6 +360,7 @@ vpxenc_tests="vpxenc_vp8_ivf
               vpxenc_vp9_webm_2pass
               vpxenc_vp9_ivf_lossless
               vpxenc_vp9_ivf_minq0_maxq0
-              vpxenc_vp9_webm_lag10_frames20"
+              vpxenc_vp9_webm_lag10_frames20
+              vpxenc_vp9_webm_non_square_par"
 
 run_tests vpxenc_verify_environment "${vpxenc_tests}"
diff --git a/tools_common.h b/tools_common.h
@@ -89,6 +89,7 @@ struct VpxInputContext {
   enum VideoFileType file_type;
   uint32_t width;
   uint32_t height;
+  struct VpxRational pixel_aspect_ratio;
   vpx_img_fmt_t fmt;
   vpx_bit_depth_t bit_depth;
   int only_i420;

diff --git a/vpxenc.c b/vpxenc.c
@@ -944,6 +944,10 @@ static void open_input_file(struct VpxInputContext *input) {
     rewind(input->file);
   }
 
+  /* Default to 1:1 pixel aspect ratio. */
+  input->pixel_aspect_ratio.numerator = 1;
+  input->pixel_aspect_ratio.denominator = 1;
+
   /* For RAW input sources, these bytes will applied on the first frame
    *  in read_frame().
    */
@@ -957,6 +961,8 @@ static void open_input_file(struct VpxInputContext *input) {
       input->file_type = FILE_TYPE_Y4M;
       input->width = input->y4m.pic_w;
       input->height = input->y4m.pic_h;
+      input->pixel_aspect_ratio.numerator = input->y4m.par_n;
+      input->pixel_aspect_ratio.denominator = input->y4m.par_d;
       input->framerate.numerator = input->y4m.fps_n;
       input->framerate.denominator = input->y4m.fps_d;
       input->fmt = input->y4m.vpx_fmt;
@@ -1390,7 +1396,8 @@ static void show_stream_config(struct stream_state *stream,
 
 
 static void open_output_file(struct stream_state *stream,
-                             struct VpxEncoderConfig *global) {
+                             struct VpxEncoderConfig *global,
+                             const struct VpxRational *pixel_aspect_ratio) {
   const char *fn = stream->config.out_fn;
   const struct vpx_codec_enc_cfg *const cfg = &stream->config.cfg;
 
@@ -1411,7 +1418,8 @@ static void open_output_file(struct stream_state *stream,
     write_webm_file_header(&stream->ebml, cfg,
                            &global->framerate,
                            stream->config.stereo_fmt,
-                           global->codec->fourcc);
+                           global->codec->fourcc,
+                           pixel_aspect_ratio);
   }
 #endif
 
@@ -2044,7 +2052,8 @@ int main(int argc, const char **argv_) {
     }
 
     FOREACH_STREAM(setup_pass(stream, &global, pass));
-    FOREACH_STREAM(open_output_file(stream, &global));
+    FOREACH_STREAM(open_output_file(stream, &global,
+                                    &input.pixel_aspect_ratio));
     FOREACH_STREAM(initialize_encoder(stream, &global));
 
 #if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH

diff --git a/webmenc.cc b/webmenc.cc
@@ -24,7 +24,8 @@ void write_webm_file_header(struct EbmlGlobal *glob,
                             const vpx_codec_enc_cfg_t *cfg,
                             const struct vpx_rational *fps,
                             stereo_format_t stereo_fmt,
-                            unsigned int fourcc) {
+                            unsigned int fourcc,
+                            const struct VpxRational *par) {
   mkvmuxer::MkvWriter *const writer = new mkvmuxer::MkvWriter(glob->stream);
   mkvmuxer::Segment *const segment = new mkvmuxer::Segment();
   segment->Init(writer);
@@ -49,6 +50,15 @@ void write_webm_file_header(struct EbmlGlobal *glob,
           segment->GetTrackByNumber(video_track_id));
   video_track->SetStereoMode(stereo_fmt);
   video_track->set_codec_id(fourcc == VP8_FOURCC ? "V_VP8" : "V_VP9");
+  if (par->numerator > 1 || par->denominator > 1) {
+    // TODO(fgalligan): Add support of DisplayUnit, Display Aspect Ratio type
+    // to WebM format.
+    const uint64_t display_width =
+        static_cast<uint64_t>(((cfg->g_w * par->numerator * 1.0) /
+                               par->denominator) + .5);
+    video_track->set_display_width(display_width);
+    video_track->set_display_height(cfg->g_h);
+  }
   if (glob->debug) {
     video_track->set_uid(kDebugTrackUid);
   }

diff --git a/webmenc.h b/webmenc.h
@@ -42,7 +42,8 @@ void write_webm_file_header(struct EbmlGlobal *glob,
                             const vpx_codec_enc_cfg_t *cfg,
                             const struct vpx_rational *fps,
                             stereo_format_t stereo_fmt,
-                            unsigned int fourcc);
+                            unsigned int fourcc,
+                            const struct VpxRational *par);
 
 void write_webm_block(struct EbmlGlobal *glob,
                       const vpx_codec_enc_cfg_t *cfg,