jpegparse: add support interlaced mjpeg

Interlaced MJPEG is a big hack. Most of the streams we've found are from old
AVID tools. There are two methods to detect interlaced stream: the container
offers a height bigger (or double) than the image's height in SOF. The other
is from a APP0 marker.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5838>
This commit is contained in:
Víctor Manuel Jáquez Leal 2023-12-19 10:39:31 +01:00
parent b39b178dd7
commit 20b636ee9a
3 changed files with 141 additions and 23 deletions
subprojects/gst-plugins-bad
gst/jpegformat
tests/check/elements

View file

@ -61,7 +61,6 @@
#include <gst/base/gstbytereader.h> #include <gst/base/gstbytereader.h>
#include <gst/codecparsers/gstjpegparser.h> #include <gst/codecparsers/gstjpegparser.h>
#include <gst/tag/tag.h> #include <gst/tag/tag.h>
#include <gst/video/video.h>
#include "gstjpegparse.h" #include "gstjpegparse.h"
@ -191,17 +190,60 @@ gst_jpeg_parse_init (GstJpegParse * parse)
parse->sof = -1; parse->sof = -1;
} }
static void
parse_avid (GstJpegParse * parse, const guint8 * data, guint16 len)
{
parse->avid = 1;
if (len > 14 && data[12] == 1) /* 1 - NTSC */
parse->field_order = GST_VIDEO_FIELD_ORDER_BOTTOM_FIELD_FIRST;
if (len > 14 && data[12] == 2) /* 2 - PAL */
parse->field_order = GST_VIDEO_FIELD_ORDER_TOP_FIELD_FIRST;
GST_INFO_OBJECT (parse, "AVID: %s",
gst_video_field_order_to_string (parse->field_order));
}
static gboolean static gboolean
gst_jpeg_parse_set_sink_caps (GstBaseParse * bparse, GstCaps * caps) gst_jpeg_parse_set_sink_caps (GstBaseParse * bparse, GstCaps * caps)
{ {
GstJpegParse *parse = GST_JPEG_PARSE_CAST (bparse); GstJpegParse *parse = GST_JPEG_PARSE_CAST (bparse);
GstStructure *s = gst_caps_get_structure (caps, 0); GstStructure *s = gst_caps_get_structure (caps, 0);
const GValue *codec_data;
const char *interlace_mode, *field_order;
GST_DEBUG_OBJECT (parse, "get sink caps %" GST_PTR_FORMAT, caps); GST_DEBUG_OBJECT (parse, "get sink caps %" GST_PTR_FORMAT, caps);
gst_structure_get_fraction (s, "framerate", gst_structure_get_fraction (s, "framerate",
&parse->framerate_numerator, &parse->framerate_denominator); &parse->framerate_numerator, &parse->framerate_denominator);
gst_structure_get_int (s, "height", &parse->orig_height);
gst_structure_get_int (s, "width", &parse->orig_width);
codec_data = gst_structure_get_value (s, "codec_data");
if (codec_data && G_VALUE_TYPE (codec_data) == GST_TYPE_BUFFER) {
GstMapInfo map;
gst_clear_buffer (&parse->codec_data);
parse->codec_data = GST_BUFFER (g_value_dup_boxed (codec_data));
if (gst_buffer_map (parse->codec_data, &map, GST_MAP_READ)) {
if (map.size > 8 && map.data[0] == 0x2c && map.data[4] == 0x18)
parse_avid (parse, map.data, map.size);
gst_buffer_unmap (parse->codec_data, &map);
}
}
interlace_mode = gst_structure_get_string (s, "interlace-mode");
if (interlace_mode) {
parse->interlace_mode =
gst_video_interlace_mode_from_string (interlace_mode);
}
if (parse->interlace_mode != GST_VIDEO_INTERLACE_MODE_PROGRESSIVE) {
field_order = gst_structure_get_string (s, "field-order");
if (field_order)
parse->field_order = gst_video_field_order_from_string (field_order);
}
return TRUE; return TRUE;
} }
@ -278,9 +320,6 @@ gst_jpeg_parse_sof (GstJpegParse * parse, GstJpegSegment * seg)
return FALSE; return FALSE;
} }
parse->width = hdr.width;
parse->height = hdr.height;
parse->colorspace = GST_JPEG_COLORSPACE_NONE; parse->colorspace = GST_JPEG_COLORSPACE_NONE;
parse->sampling = GST_JPEG_SAMPLING_NONE; parse->sampling = GST_JPEG_SAMPLING_NONE;
@ -346,6 +385,24 @@ gst_jpeg_parse_sof (GstJpegParse * parse, GstJpegSegment * seg)
break; break;
} }
if (hdr.width != parse->width || hdr.height != parse->height) {
parse->width = hdr.width;
parse->height = hdr.height;
if (parse->first_picture && !parse->multiscope) {
if (parse->orig_height > 0
&& parse->height < ((parse->orig_height * 3) / 4)) {
parse->interlace_mode = GST_VIDEO_INTERLACE_MODE_INTERLEAVED;
} else if (parse->avid) {
if (parse->orig_height == 0)
parse->orig_height = 2 * hdr.height;
parse->interlace_mode = GST_VIDEO_INTERLACE_MODE_INTERLEAVED;
}
}
parse->first_picture = FALSE;
}
GST_INFO_OBJECT (parse, "SOF [%dx%d] %d comp - %s", parse->width, GST_INFO_OBJECT (parse, "SOF [%dx%d] %d comp - %s", parse->width,
parse->height, hdr.num_components, parse->height, hdr.num_components,
GST_STR_NULL (colorspace_to_string (parse->colorspace))); GST_STR_NULL (colorspace_to_string (parse->colorspace)));
@ -443,6 +500,9 @@ gst_jpeg_parse_app0 (GstJpegParse * parse, GstJpegSegment * seg)
if (!gst_byte_reader_get_uint8 (&reader, &unit)) if (!gst_byte_reader_get_uint8 (&reader, &unit))
return FALSE; return FALSE;
parse->avid = TRUE;
parse->field = unit == 1 ? 0 : 1;
/* TODO: update caps for interlaced MJPEG */ /* TODO: update caps for interlaced MJPEG */
GST_DEBUG_OBJECT (parse, "MJPEG interleaved field: %d", unit); GST_DEBUG_OBJECT (parse, "MJPEG interleaved field: %d", unit);
@ -580,39 +640,50 @@ gst_jpeg_parse_com (GstJpegParse * parse, GstJpegSegment * seg)
GstByteReader reader; GstByteReader reader;
const guint8 *data = NULL; const guint8 *data = NULL;
guint16 size; guint16 size;
gchar *comment; const gchar *buf;
gst_byte_reader_init (&reader, seg->data + seg->offset, seg->size); gst_byte_reader_init (&reader, seg->data + seg->offset, seg->size);
gst_byte_reader_skip_unchecked (&reader, 2); gst_byte_reader_skip_unchecked (&reader, 2);
size = gst_byte_reader_get_remaining (&reader); size = gst_byte_reader_get_remaining (&reader);
if (!gst_byte_reader_get_data (&reader, size, &data)) if (!gst_byte_reader_get_data (&reader, size, &data))
return FALSE; return FALSE;
comment = get_utf8_from_data (data, size); buf = (const gchar *) data;
if (!comment) /* buggy avid, it puts EOI only at every 10th frame */
return FALSE; if (g_str_has_prefix (buf, "AVID")) {
parse_avid (parse, data, size);
} else if (g_str_has_prefix (buf, "MULTISCOPE II")) {
parse->x_density = 1;
parse->y_density = 2;
parse->multiscope = TRUE;
} else {
gchar *comment;
GST_INFO_OBJECT (parse, "comment found: %s", comment); comment = get_utf8_from_data (data, size);
gst_tag_list_add (get_tag_list (parse), GST_TAG_MERGE_REPLACE, if (!comment)
GST_TAG_COMMENT, comment, NULL); return FALSE;
g_free (comment);
GST_INFO_OBJECT (parse, "comment found: %s", comment);
gst_tag_list_add (get_tag_list (parse), GST_TAG_MERGE_REPLACE,
GST_TAG_COMMENT, comment, NULL);
g_free (comment);
}
return TRUE; return TRUE;
} }
/* reset per image */
static void static void
gst_jpeg_parse_reset (GstJpegParse * parse) gst_jpeg_parse_reset (GstJpegParse * parse)
{ {
parse->width = 0;
parse->height = 0;
parse->last_offset = 0; parse->last_offset = 0;
parse->state = 0; parse->state = 0;
parse->sof = -1; parse->sof = -1;
parse->adobe_transform = 0; parse->adobe_transform = 0;
parse->x_density = 0; parse->x_density = 0;
parse->y_density = 0; parse->y_density = 0;
parse->field = 0;
if (parse->tags) { if (parse->tags) {
gst_tag_list_unref (parse->tags); gst_tag_list_unref (parse->tags);
@ -637,7 +708,9 @@ gst_jpeg_parse_set_new_caps (GstJpegParse * parse)
if (parse->width > 0) if (parse->width > 0)
gst_caps_set_simple (caps, "width", G_TYPE_INT, parse->width, NULL); gst_caps_set_simple (caps, "width", G_TYPE_INT, parse->width, NULL);
if (parse->width > 0) if (parse->orig_height > 0 && parse->orig_height > parse->height)
gst_caps_set_simple (caps, "height", G_TYPE_INT, parse->orig_height, NULL);
else if (parse->height > 0)
gst_caps_set_simple (caps, "height", G_TYPE_INT, parse->height, NULL); gst_caps_set_simple (caps, "height", G_TYPE_INT, parse->height, NULL);
if (parse->sof >= 0) if (parse->sof >= 0)
gst_caps_set_simple (caps, "sof-marker", G_TYPE_INT, parse->sof, NULL); gst_caps_set_simple (caps, "sof-marker", G_TYPE_INT, parse->sof, NULL);
@ -650,6 +723,14 @@ gst_jpeg_parse_set_new_caps (GstJpegParse * parse)
sampling_to_string (parse->sampling), NULL); sampling_to_string (parse->sampling), NULL);
} }
gst_caps_set_simple (caps, "interlace-mode", G_TYPE_STRING,
gst_video_interlace_mode_to_string (parse->interlace_mode), NULL);
if (parse->interlace_mode == GST_VIDEO_INTERLACE_MODE_INTERLEAVED) {
gst_caps_set_simple (caps, "field-order", G_TYPE_STRING,
gst_video_field_order_to_string (parse->field_order), NULL);
}
gst_caps_set_simple (caps, "framerate", GST_TYPE_FRACTION, gst_caps_set_simple (caps, "framerate", GST_TYPE_FRACTION,
parse->framerate_numerator, parse->framerate_denominator, NULL); parse->framerate_numerator, parse->framerate_denominator, NULL);
@ -658,6 +739,11 @@ gst_jpeg_parse_set_new_caps (GstJpegParse * parse)
parse->x_density, parse->y_density, NULL); parse->x_density, parse->y_density, NULL);
} }
if (parse->codec_data) {
gst_caps_set_simple (caps, "codec_data", GST_TYPE_BUFFER, parse->codec_data,
NULL);
}
if (parse->prev_caps && gst_caps_is_equal_fixed (caps, parse->prev_caps)) { if (parse->prev_caps && gst_caps_is_equal_fixed (caps, parse->prev_caps)) {
gst_caps_unref (caps); gst_caps_unref (caps);
return TRUE; return TRUE;
@ -711,6 +797,8 @@ gst_jpeg_parse_handle_frame (GstBaseParse * bparse, GstBaseParseFrame * frame,
GstJpegSegment seg; GstJpegSegment seg;
guint offset; guint offset;
GST_TRACE_OBJECT (parse, "frame %" GST_PTR_FORMAT, frame->buffer);
if (!gst_buffer_map (frame->buffer, &mapinfo, GST_MAP_READ)) if (!gst_buffer_map (frame->buffer, &mapinfo, GST_MAP_READ))
return GST_FLOW_ERROR; return GST_FLOW_ERROR;
@ -754,7 +842,9 @@ gst_jpeg_parse_handle_frame (GstBaseParse * bparse, GstBaseParseFrame * frame,
switch (marker) { switch (marker) {
case GST_JPEG_MARKER_SOI: case GST_JPEG_MARKER_SOI:
/* This means that new SOI comes without an previous EOI. */ /* This means that new SOI comes without an previous EOI. */
if (offset > 2) { if (offset > 2
&& (parse->interlace_mode == GST_VIDEO_INTERLACE_MODE_PROGRESSIVE
|| parse->field == 0)) {
/* If already some data segment parsed, push it as a frame. */ /* If already some data segment parsed, push it as a frame. */
if (valid_state (parse->state, GST_JPEG_PARSER_STATE_GOT_SOS)) { if (valid_state (parse->state, GST_JPEG_PARSER_STATE_GOT_SOS)) {
gst_buffer_unmap (frame->buffer, &mapinfo); gst_buffer_unmap (frame->buffer, &mapinfo);
@ -785,8 +875,15 @@ gst_jpeg_parse_handle_frame (GstBaseParse * bparse, GstBaseParseFrame * frame,
parse->state |= GST_JPEG_PARSER_STATE_GOT_SOI; parse->state |= GST_JPEG_PARSER_STATE_GOT_SOI;
break; break;
case GST_JPEG_MARKER_EOI: case GST_JPEG_MARKER_EOI:
gst_buffer_unmap (frame->buffer, &mapinfo); if (parse->interlace_mode == GST_VIDEO_INTERLACE_MODE_PROGRESSIVE
return gst_jpeg_parse_finish_frame (parse, frame, seg.offset); || parse->field == 1) {
gst_buffer_unmap (frame->buffer, &mapinfo);
return gst_jpeg_parse_finish_frame (parse, frame, seg.offset);
} else if (parse->interlace_mode == GST_VIDEO_INTERLACE_MODE_INTERLEAVED
&& parse->field == 0) {
parse->field = 1;
parse->state = 0;
}
break; break;
case GST_JPEG_MARKER_SOS: case GST_JPEG_MARKER_SOS:
if (!valid_state (parse->state, GST_JPEG_PARSER_STATE_GOT_SOF)) if (!valid_state (parse->state, GST_JPEG_PARSER_STATE_GOT_SOF))
@ -874,6 +971,11 @@ gst_jpeg_parse_start (GstBaseParse * bparse)
parse->framerate_numerator = 0; parse->framerate_numerator = 0;
parse->framerate_denominator = 1; parse->framerate_denominator = 1;
parse->first_picture = TRUE;
parse->interlace_mode = GST_VIDEO_INTERLACE_MODE_PROGRESSIVE;
parse->field_order = GST_VIDEO_FIELD_ORDER_TOP_FIELD_FIRST;
gst_jpeg_parse_reset (parse); gst_jpeg_parse_reset (parse);
gst_base_parse_set_min_frame_size (bparse, 2); gst_base_parse_set_min_frame_size (bparse, 2);
@ -890,6 +992,7 @@ gst_jpeg_parse_stop (GstBaseParse * bparse)
gst_tag_list_unref (parse->tags); gst_tag_list_unref (parse->tags);
parse->tags = NULL; parse->tags = NULL;
} }
gst_clear_buffer (&parse->codec_data);
gst_clear_caps (&parse->prev_caps); gst_clear_caps (&parse->prev_caps);
return TRUE; return TRUE;

View file

@ -26,6 +26,7 @@
#include <gst/gst.h> #include <gst/gst.h>
#include <gst/base/gstbaseparse.h> #include <gst/base/gstbaseparse.h>
#include <gst/video/video.h>
G_BEGIN_DECLS G_BEGIN_DECLS
@ -50,11 +51,21 @@ struct _GstJpegParse {
guint last_offset; guint last_offset;
gint state; gint state;
gboolean first_picture;
gboolean multiscope;
gboolean avid;
gint8 sof; gint8 sof;
gint8 adobe_transform; gint8 adobe_transform;
/* the parsed frame size */ /* the parsed frame size */
guint16 width, height; guint16 width, height;
gint orig_width, orig_height;
GstBuffer *codec_data;
GstVideoInterlaceMode interlace_mode;
GstVideoFieldOrder field_order;
guint field;
/* format color space */ /* format color space */
guint colorspace; guint colorspace;

View file

@ -192,7 +192,8 @@ GST_START_TEST (test_parse_single_byte)
caps_in = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, FALSE, caps_in = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, FALSE,
NULL); NULL);
caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE, caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE,
"framerate", GST_TYPE_FRACTION, 0, 1, NULL); "framerate", GST_TYPE_FRACTION, 0, 1, "interlace-mode", G_TYPE_STRING,
"progressive", NULL);
/* Push the data byte by byte, injecting some garbage. */ /* Push the data byte by byte, injecting some garbage. */
buffer_in = make_buffers_in (buffer_in, test_data_garbage); buffer_in = make_buffers_in (buffer_in, test_data_garbage);
@ -262,7 +263,8 @@ GST_START_TEST (test_parse_all_in_one_buf)
buffer_in = g_list_append (buffer_in, buffer); buffer_in = g_list_append (buffer_in, buffer);
caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE, caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE,
"framerate", GST_TYPE_FRACTION, 0, 1, NULL); "framerate", GST_TYPE_FRACTION, 0, 1, "interlace-mode", G_TYPE_STRING,
"progressive", NULL);
buffer_out = make_buffers_out (buffer_out, test_data_short_frame); buffer_out = make_buffers_out (buffer_out, test_data_short_frame);
buffer_out = make_buffers_out (buffer_out, test_data_normal_frame); buffer_out = make_buffers_out (buffer_out, test_data_normal_frame);
buffer_out = make_buffers_out (buffer_out, test_data_entropy); buffer_out = make_buffers_out (buffer_out, test_data_entropy);
@ -328,7 +330,8 @@ GST_START_TEST (test_parse_app1_exif)
caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE, caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE,
"framerate", GST_TYPE_FRACTION, 0, 1, "width", G_TYPE_INT, 80, "height", "framerate", GST_TYPE_FRACTION, 0, 1, "width", G_TYPE_INT, 80, "height",
G_TYPE_INT, 60, "sof-marker", G_TYPE_INT, 0, "colorspace", G_TYPE_STRING, G_TYPE_INT, 60, "sof-marker", G_TYPE_INT, 0, "colorspace", G_TYPE_STRING,
"sYUV", "sampling", G_TYPE_STRING, "YCbCr-4:2:0", NULL); "sYUV", "sampling", G_TYPE_STRING, "YCbCr-4:2:0", "interlace-mode",
G_TYPE_STRING, "progressive", NULL);
buffer_in = make_my_input_buffer (test_data_app1_exif, buffer_in = make_my_input_buffer (test_data_app1_exif,
sizeof (test_data_app1_exif)); sizeof (test_data_app1_exif));
@ -354,7 +357,8 @@ GST_START_TEST (test_parse_comment)
caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE, caps_out = gst_caps_new_simple ("image/jpeg", "parsed", G_TYPE_BOOLEAN, TRUE,
"framerate", GST_TYPE_FRACTION, 0, 1, "width", G_TYPE_INT, 80, "height", "framerate", GST_TYPE_FRACTION, 0, 1, "width", G_TYPE_INT, 80, "height",
G_TYPE_INT, 60, "sof-marker", G_TYPE_INT, 0, "colorspace", G_TYPE_STRING, G_TYPE_INT, 60, "sof-marker", G_TYPE_INT, 0, "colorspace", G_TYPE_STRING,
"sYUV", "sampling", G_TYPE_STRING, "YCbCr-4:2:0", NULL); "sYUV", "sampling", G_TYPE_STRING, "YCbCr-4:2:0", "interlace-mode",
G_TYPE_STRING, "progressive", NULL);
buffer_in = make_my_input_buffer (test_data_comment, buffer_in = make_my_input_buffer (test_data_comment,
sizeof (test_data_comment)); sizeof (test_data_comment));