Deinterlacing support

avpipe.go

@@ -307,6 +307,7 @@ type XcParams struct {
 	Rotate                 int         `json:"rotate"`
 	Profile                string      `json:"profile"`
 	Level                  int         `json:"level"`
+        Deinterlace            int         `json:"deinterlace"`
 }
 
 // NewXcParams initializes a XcParams struct with unset/default values
@@ -1402,6 +1403,7 @@ func getCParams(params *XcParams) (*C.xcparams_t, error) {
 		rotate:                    C.int(params.Rotate),
 		profile:                   C.CString(params.Profile),
 		level:                     C.int(params.Level),
+                deinterlace:               C.dif_type(params.Deinterlace),
 
 		// All boolean params are handled below
 	}

doc/dev.md

@@ -0,0 +1,38 @@
+# DEVELOPER NOTES
+
+
+
+## HOW TO
+
+### Transcode a "content part" into ABR segments
+
+```
+./bin/exc -f hqp_3VQ52.mp4 -format dash -xc-type video -video-seg-duration-ts 120120
+```
+
+- output files: `./O/O1/`
+
+
+### Transcode a "content part" into encrypted ABR segments
+
+
+```
+./bin/exc -f hqp_3VQ52.mp4 -format dash -xc-type video -video-seg-duration-ts 120120 -crypt-scheme cenc -crypt-key 13c396f04c947a0b0c9794f7f114a614 -crypt-kid 351a81dbaf704ab43ce7f3b4fbc603b3 -crypt-iv 9c33afa0bff3c27f671fa8e91f31f2c9
+```
+
+- output files: `./O/O1/`
+
+
+## Special Use Cases
+
+### Deinterlacing
+
+Deinterlacing using the bwdiff "field" filter creates two frames for each input frame and requires adjusting the framerate.
+When doubling the framerate the timebase (timescale) may not accommodate the new frame duration and it needs to be adjusted as well.
+
+Deinterlace a source file (25fps) using CRF 16 and max_rate 20 Mbps.  Note video timebase changes to 50 to accommodate new frame duration - this will be adjusted by ffmpeg to 12,800.  The resulting frame duration is 265  (12800/50).
+
+```
+  ./bin/exc -f TestFile.mxf -format fmp4-segment -xc-type video  -crf 16 -seg-duration 30.000 -force-keyint 100 -enc-height 1080 -enc-width 1920 -deinterlace 1  -video-time-base 50 -video-frame-duration-ts 256 -rc-max-rate 200000000 -rc-buffer-size 40000000
+```
+

doc/transcoding.md

@@ -0,0 +1,56 @@
+# AVPIPE TRANSCODING
+
+
+## Job Types
+
+The avpipe library is designed to accommodate specific transcoding operations.
+
+### Mez creation - VOD
+
+The input file can be any format and any size - this is the source media for the mezzanine.
+
+The operation creates 30 second (approx) content parts, fragmented MP4, encoded at the desired resolution and bitrate, and conditioned with GOP and keyframe interval of approx 2 sec.
+
+
+### Mez creation - Live
+
+The input is a live stream RTMP, MPEGTS, SRT or HLS.
+The operation creates 30 second (approx) content parts, fragmented MP4, encoded at the desired resolution and bitrate, and conditioned with GOP and keyframe interval of approx 2 sec.
+
+### ABR segments
+
+The input is a conditioned mez content part, as created by the mez creation operation
+
+The output is a set of "CMAF" segments (2 sec approx) and the corresponding 'init' segment
+
+### Mux-only File Maker
+
+> TODO
+
+### File Maker
+
+> TODO
+
+### Thumbnail Extraction
+
+The input is a conditioned mez content part, as created by the mez creation operation.
+
+The output is a set of individual frames.
+
+## Common Use Cases
+
+> TODO
+
+## Special Use Cases
+
+### Deinterlacing
+
+Deinterlacing using the bwdif "field" filter creates two frames for each input frame and requires adjusting the framerate.
+
+When doubling the framerate the timebase (timescale) may not accommodate the new frame duration and it needs to be adjusted as well.
+
+For example if the input frame rate is 25 and timebase is 1/25, the input frame duration is 1.  When deinterlacing using the bwdif "field" filter
+we need to change the output timebase to 1/50 and the frame rate becomes 50 fps.
+
+Note that ffmpeg will force the timebase denominator to be greater than 10,000 and will double our "1/50" timebase until it becomes 1/12800.  The new frame duration,
+for 50 fps, will be 256.

elvxc/cmd/transcode.go

@@ -336,6 +336,7 @@ func InitTranscode(cmdRoot *cobra.Command) error {
 	cmdTranscode.PersistentFlags().Int32("rotate", 0, "Rotate the output video frame (valid values 0, 90, 180, 270).")
 	cmdTranscode.PersistentFlags().StringP("profile", "", "", "Encoding profile for video. If it is not determined, it will be set automatically.")
 	cmdTranscode.PersistentFlags().Int32("level", 0, "Encoding level for video. If it is not determined, it will be set automatically.")
+	cmdTranscode.PersistentFlags().Int32("deinterlace", 0, "Deinterlace filter (values 0 - none, 1 - bwdif_field, 2 - bwdif_frame send_frame).")
 
 	return nil
 }
@@ -621,6 +622,11 @@ func doTranscode(cmd *cobra.Command, args []string) error {
 
 	profile := cmd.Flag("profile").Value.String()
 
+	deinterlace, err := cmd.Flags().GetInt32("deinterlace")
+	if err != nil {
+		return fmt.Errorf("Invalid deinterlace value")
+	}
+
 	cryptScheme := avpipe.CryptNone
 	val := cmd.Flag("crypt-scheme").Value.String()
 	if len(val) > 0 {
@@ -719,6 +725,7 @@ func doTranscode(cmd *cobra.Command, args []string) error {
 		Rotate:                 int(rotate),
 		Profile:                profile,
 		Level:                  int(level),
+		Deinterlace:            int(deinterlace),
 	}
 
 	err = getAudioIndexes(params, audioIndex)

exc/elv_xc.c

@@ -361,6 +361,10 @@ in_stat(
         if (debug_frame_level)
             elv_dbg("IN STAT stream_index=%d, fd=%d, video frame read=%"PRId64, stream_index, fd, c->video_frames_read);
         break;
+    case in_stat_first_keyframe_pts:
+        if (debug_frame_level)
+            elv_dbg("IN STAT fd=%d, first keyframe PTS=%"PRId64", url=%s", fd, c->first_key_frame_pts, c->url);
+        break;
     case in_stat_data_scte35:
         if (debug_frame_level)
             elv_dbg("IN STAT stream_index=%d, fd=%d, data=%s", stream_index, fd, c->data);
@@ -1028,6 +1032,7 @@ usage(
         "\t-d :                     (optional) Decoder name. For video default is \"h264\", can be: \"h264\", \"h264_cuvid\", \"jpeg2000\", \"hevc\"\n"
         "\t                                    For audio default is \"aac\", but for ts files should be set to \"ac3\"\n"
         "\t-debug-frame-level :     (optional) Enable/disable debug frame level. Default is 0, must be 0 or 1.\n"
+        "\t-deinterlace :           (optional) Deinterlace filter. Default is 0 (none), can be: 1 (bwdif send_field), 2 (bwdif send_frame)\n"
         "\t-duration-ts :           (optional) Default: -1 (entire stream)\n"
         "\t-e :                     (optional) Video encoder name. Default is \"libx264\", can be: \"libx264\", \"libx265\", \"h264_nvenc\", \"hevc_nvenc\", \"h264_videotoolbox\", or \"mjpeg\"\n"
         "\t-enc-height :            (optional) Default: -1 (use source height)\n"
@@ -1167,6 +1172,7 @@ main(
         .start_fragment_index = 0,          /* Default is zero */
         .sync_audio_to_stream_id = -1,      /* Default -1 (no sync to a video stream) */
         .rotate = 0,                        /* Default 0 (means no transpose/rotation) */
+        .deinterlace = 0,                   /* Default 0 (no deinterlacing) */
         .xc_type = xc_none,
         .video_bitrate = -1,                /* not used if using CRF */
         .watermark_text = NULL,
@@ -1284,7 +1290,12 @@ main(
                 if (p.debug_frame_level != 0 && p.debug_frame_level != 1) {
                     usage(argv[0], argv[i], EXIT_FAILURE);
                 }
-            } else if (strlen(argv[i]) > 2) {
+            } else if (!strcmp(argv[i], "-deinterlace")) {
+                if (sscanf(argv[i+1], "%d", &p.deinterlace) != 1) {
+                    usage(argv[0], argv[i], EXIT_FAILURE);
+                }
+            }
+            else if (strlen(argv[i]) > 2) {
                 usage(argv[0], argv[i], EXIT_FAILURE);
             } else {
                 p.dcodec = strdup(argv[i+1]);

libavpipe/include/avpipe_xc.h

@@ -371,6 +371,13 @@ typedef enum image_type {
     gif_image
 } image_type;
 
+// deinterlacing filter types
+typedef enum dif_type {
+    dif_none        = 0, // No deinterlacing
+    dif_bwdif       = 1, // Use filter bwdif mode 'send_field' (two frames per input frame)
+    dif_bwdif_frame = 2  // Use filter bwdif mode 'send_frame' (one frame per input frame)
+} dif_type;
+
 #define DRAW_TEXT_SHADOW_OFFSET     0.075
 #define MAX_EXTRACT_IMAGES_SZ       100
 
@@ -454,6 +461,7 @@ typedef struct xcparams_t {
     int         rotate;                     // For video transpose or rotation
     char        *profile;
     int         level;
+    dif_type    deinterlace;                // Deinterlacing filter
 } xcparams_t;
 
 #define MAX_CODEC_NAME  256

libavpipe/src/avpipe_xc.c

@@ -1182,6 +1182,7 @@ prepare_video_encoder(
             out_stream->time_base = (AVRational) {1, params->video_time_base};
         else
             out_stream->time_base = in_stream->time_base;
+
         out_stream->avg_frame_rate = decoder_context->format_context->streams[decoder_context->video_stream_index]->avg_frame_rate;
         out_stream->codecpar->codec_tag = 0;
 
@@ -1237,13 +1238,15 @@ prepare_video_encoder(
         encoder_codec_context->time_base = (AVRational) {1, params->video_time_base};
     else
         encoder_codec_context->time_base = decoder_context->codec_context[index]->time_base;
+
     encoder_codec_context->sample_aspect_ratio = decoder_context->codec_context[index]->sample_aspect_ratio;
     if (params->video_bitrate > 0)
         encoder_codec_context->bit_rate = params->video_bitrate;
     if (params->rc_buffer_size > 0)
         encoder_codec_context->rc_buffer_size = params->rc_buffer_size;
     if (params->rc_max_rate > 0)
         encoder_codec_context->rc_max_rate = params->rc_max_rate;
+
     encoder_codec_context->framerate = decoder_context->codec_context[index]->framerate;
 
     // This needs to be set before open (ffmpeg samples have it wrong)
@@ -3416,6 +3419,42 @@ get_filter_str(
 {
     *filter_str = NULL;
 
+    // Validate filter compatibility
+    // Note these filters can theoretically be made to work together but not a real use case
+    if (params->rotate > 0 || params->deinterlace != dif_none) {
+        if ((params->watermark_text && *params->watermark_text != '\0') ||
+            (params->watermark_overlay && params->watermark_overlay[0] != '\0')) {
+            elv_err("Incompatible filter parameters - watermark not supported with rotate and deinterlacing");
+            return eav_param;
+        }
+        if (params->rotate > 0 && params->deinterlace != dif_none) {
+            elv_err("Incompatible filter parameters - both rotate and deinterlacing");
+            return eav_param;
+        }
+        if (params->deinterlace == dif_bwdif) {
+            // This filter needs to create two output frames for each input frame and
+            // requires the caller to specify the new frame duration (1/2 of input frame duration)
+            if (params->video_frame_duration_ts == 0) {
+                elv_err("Incorrect filter spec - deinterlacing requires frame duration");
+                return eav_param;
+            }
+        }
+    }
+
+    //  General syntax for the bwdif filter:
+    //  "bwdif=mode=send_frame:parity=auto:deint=all"
+    switch (params->deinterlace) {
+        case dif_bwdif:
+            *filter_str = strdup("bwdif=mode=send_field");
+            return eav_success;
+        case dif_bwdif_frame:
+            *filter_str = strdup("bwdif=mode=send_frame");
+            return eav_success;
+        default:
+            // Nothing to do
+            break;
+    }
+
     if (params->rotate > 0) {
         switch (params->rotate) {
         case 90:
@@ -4633,7 +4672,8 @@ log_params(
         "video_frame_duration_ts=%d "
         "rotate=%d "
         "profile=%s "
-        "level=%d",
+        "level=%d "
+        "deinterlace=%d",
         params->stream_id, params->url,
         avpipe_version(),
         params->bypass_transcoding, params->skip_decoding,
@@ -4657,7 +4697,7 @@ log_params(
         params->filter_descriptor,
         params->extract_image_interval_ts, params->extract_images_sz,
         1, params->video_time_base, params->video_frame_duration_ts, params->rotate,
-        params->profile ? params->profile : "", params->level);
+        params->profile ? params->profile : "", params->level,  params->deinterlace);
     elv_log("AVPIPE XCPARAMS %s", buf);
 }