FaSTTUBe/GST-Tensordecoder-ov_ep
61
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Cope with short startcodes in the h264 bytestream

Browse Source
This commit is contained in:
Sjoerd Simons 2010-06-02 19:16:20 +01:00 committed by Wim Taymans
parent 3486f49808
commit c39e82a1ce
1 changed files with 23 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
gst/rtp/gstrtph264pay.c
View File

@ -485,8 +485,8 @@ next_start_code (guint8 * data, guint size)
* sense because our search 'alphabet' is binary - 0 & 1 only. * sense because our search 'alphabet' is binary - 0 & 1 only.
* This allow us to simplify the general BM algorithm to a very * This allow us to simplify the general BM algorithm to a very
* simple form. */ * simple form. */
/* assume 1 is in the 4th byte */ /* assume 1 is in the 3th byte */
guint offset = 3; guint offset = 2;
while (offset < size) { while (offset < size) {
if (1 == data[offset]) { if (1 == data[offset]) {
@ -494,23 +494,21 @@ next_start_code (guint8 * data, guint size)
if (0 == data[--shift]) { if (0 == data[--shift]) {
if (0 == data[--shift]) { if (0 == data[--shift]) {
if (0 == data[--shift]) { return shift;
return shift;
}
} }
} }
/* The jump is always 4 because of the 1 previously matched. /* The jump is always 3 because of the 1 previously matched.
* All the 0's must be after this '1' matched at offset */ * All the 0's must be after this '1' matched at offset */
offset += 4; offset += 3;
} else if (0 == data[offset]) { } else if (0 == data[offset]) {
/* maybe next byte is 1? */ /* maybe next byte is 1? */
offset++; offset++;
} else { } else {
/* can jump 4 bytes forward */ /* can jump 3 bytes forward */
offset += 4; offset += 3;
} }
/* at each iteration, we rescan in a backward manner until /* at each iteration, we rescan in a backward manner until
* we match 0.0.0.1 in reverse order. Since our search string * we match 0.0.1 in reverse order. Since our search string
* has only 2 'alpabets' (i.e. 0 & 1), we know that any * has only 2 'alpabets' (i.e. 0 & 1), we know that any
* mismatch will force us to shift a fixed number of steps */ * mismatch will force us to shift a fixed number of steps */
} }
@ -985,8 +983,8 @@ gst_rtp_h264_pay_handle_buffer (GstBaseRTPPayload * basepayload,
/* first pass to locate NALs and parse SPS/PPS */ /* first pass to locate NALs and parse SPS/PPS */
while (size > 4) { while (size > 4) {
/* skip start code */ /* skip start code */
data += 4; data += 3;
size -= 4; size -= 3;
if (rtph264pay->scan_mode == GST_H264_SCAN_MODE_SINGLE_NAL) { if (rtph264pay->scan_mode == GST_H264_SCAN_MODE_SINGLE_NAL) {
/* we are told that there is only a single NAL in this packet so that we /* we are told that there is only a single NAL in this packet so that we
@ -1048,13 +1046,24 @@ gst_rtp_h264_pay_handle_buffer (GstBaseRTPPayload * basepayload,
/* second pass to payload and push */ /* second pass to payload and push */
data = nal_data; data = nal_data;
for (i = 0; i < nal_queue->len; i++) { for (i = 0; i < nal_queue->len; i++) {
guint size;
nal_len = g_array_index (nal_queue, guint, i); nal_len = g_array_index (nal_queue, guint, i);
/* skip start code */ /* skip start code */
data += 4; data += 3;
/* Trim the end unless we're the last NAL in the buffer.
* In case we're not at the end of the buffer we know the next block
* starts with 0x000001 so all the 0x00 bytes at the end of this one are
* trailing 0x0 that can be discarded */
size = nal_len;
if (i + 1 != nal_queue->len)
for ( ; size > 1 && data[size - 1] == 0x0; size--)
/* skip */;
/* put the data in one or more RTP packets */ /* put the data in one or more RTP packets */
ret = ret =
gst_rtp_h264_pay_payload_nal (basepayload, data, nal_len, timestamp, gst_rtp_h264_pay_payload_nal (basepayload, data, size, timestamp,
buffer); buffer);
if (ret != GST_FLOW_OK) { if (ret != GST_FLOW_OK) {
break; break;