场景
指定海康摄像机的分辨率是1920*1080P，I帧间隔是25，视频帧率是25fps。查看AVFormatContext结构体中streams数组中的第一个元素codec
codec_id=AV_CODEC_ID_H264, width=1920, height=1080, coded_with=1920, coded_height=1088, pix_fmt=AV_PIX_FMT_YUVJ420P
疑问
coded_height为什么不是1080
解答
当前SPS的帧的宽 = (sps_info.pic_width_in_mbs_minus1 + 1) * 16
当前SPS的帧的高 = (sps_info.pic_height_in_map_units_minus1 + 1) * 16
通过如下代码计算得到的宽高（1888 x 1920）却不完全等同于源视频的宽高（1920 x 1080）。
根据编码规则计算到的宽高是1920x1088，而MP4中读取的是1920x1080。那么，如何FFmpeg是如何修正SPS里的宽高计算呢？
实际上，SPS还有crop系列字段，由crop表示帧是否被裁剪、crop_left、crop_right、crop_top和crop_bottom表示要裁剪的值得到正确的宽高值。
相关知识
sps和pps的结构参考：h264编码 里面的Sequence parameter set RBSP syntax
0x00000001或者0x000001是起始码，0x67是sps的开头，0x68是pps的开头。
0x42代表profile_idc,后面八位是constraint_set0_flag和reserved_zero_4bits，都设为0，0x0a是level_idc,
接着后面为图方便能用0表示的都用了。这里要注意是ue(v)表示该域是可变位，使用的指数-哥伦布编码
我的目的主要是设置正确帧高度和帧高度，所以只要填充 pic_width_in_mbs_minus1和 pic_height_in_map_units_minus1，
将它们的十六进制数写入sps_pps，
如果宽度和高度不是16的倍数，要填frame_cropping_flag
参考
http://m.blog.csdn.net/stn_lcd/article/details/74390636
http://m.blog.csdn.net/leixiaohua1020/article/details/45001033
static int h264_init_ps(H264Context *h, const H264SliceContext *sl, int first_slice)
{
const SPS *sps;
int needs_reinit = 0, must_reinit, ret;
if (first_slice) {
av_buffer_unref(&h->ps.pps_ref);
h->ps.pps = NULL;
h->ps.pps_ref = av_buffer_ref(h->ps.pps_list[sl->pps_id]);
if (!h->ps.pps_ref)
return AVERROR(ENOMEM);
h->ps.pps = (const PPS*)h->ps.pps_ref->data;
}
if (h->ps.sps != (const SPS*)h->ps.sps_list[h->ps.pps->sps_id]->data) {
av_buffer_unref(&h->ps.sps_ref);
h->ps.sps = NULL;
h->ps.sps_ref = av_buffer_ref(h->ps.sps_list[h->ps.pps->sps_id]);
if (!h->ps.sps_ref)
return AVERROR(ENOMEM);
h->ps.sps = (const SPS*)h->ps.sps_ref->data;
if (h->mb_width != h->ps.sps->mb_width ||
h->mb_height != h->ps.sps->mb_height ||
h->cur_bit_depth_luma != h->ps.sps->bit_depth_luma ||
h->cur_chroma_format_idc != h->ps.sps->chroma_format_idc
)
needs_reinit = 1;
if (h->bit_depth_luma != h->ps.sps->bit_depth_luma ||
h->chroma_format_idc != h->ps.sps->chroma_format_idc)
needs_reinit = 1;
}
sps = h->ps.sps;
must_reinit = (h->context_initialized &&
( 16*sps->mb_width != h->avctx->coded_width
|| 16*sps->mb_height != h->avctx->coded_height
|| h->cur_bit_depth_luma != sps->bit_depth_luma
|| h->cur_chroma_format_idc != sps->chroma_format_idc
|| h->mb_width != sps->mb_width
|| h->mb_height != sps->mb_height
));
if (h->avctx->pix_fmt == AV_PIX_FMT_NONE
|| (non_j_pixfmt(h->avctx->pix_fmt) != non_j_pixfmt(get_pixel_format(h, 0))))
must_reinit = 1;
if (first_slice && av_cmp_q(sps->sar, h->avctx->sample_aspect_ratio))
must_reinit = 1;
if (!h->setup_finished) {
h->avctx->profile = ff_h264_get_profile(sps);
h->avctx->level = sps->level_idc;
h->avctx->refs = sps->ref_frame_count;
h->mb_width = sps->mb_width;
h->mb_height = sps->mb_height;
h->mb_num = h->mb_width * h->mb_height;
h->mb_stride = h->mb_width + 1;
h->b_stride = h->mb_width * 4;
h->chroma_y_shift = sps->chroma_format_idc <= 1; // 400 uses yuv420p
h->width = 16 * h->mb_width;
h->height = 16 * h->mb_height;
ret = init_dimensions(h);
if (ret < 0)
return ret;
if (sps->video_signal_type_present_flag) {
h->avctx->color_range = sps->full_range > 0 ? AVCOL_RANGE_JPEG
: AVCOL_RANGE_MPEG;
if (sps->colour_description_present_flag) {
if (h->avctx->colorspace != sps->colorspace)
needs_reinit = 1;
h->avctx->color_primaries = sps->color_primaries;
h->avctx->color_trc = sps->color_trc;
h->avctx->colorspace = sps->colorspace;
}
}
}
if (!h->context_initialized || must_reinit || needs_reinit) {
int flush_changes = h->context_initialized;
h->context_initialized = 0;
if (sl != h->slice_ctx) {
av_log(h->avctx, AV_LOG_ERROR,
"changing width %d -> %d / height %d -> %d on "
"slice %d\n",
h->width, h->avctx->coded_width,
h->height, h->avctx->coded_height,
h->current_slice + 1);
return AVERROR_INVALIDDATA;
}
av_assert1(first_slice);
if (flush_changes)
ff_h264_flush_change(h);
if ((ret = get_pixel_format(h, 1)) < 0)
return ret;
h->avctx->pix_fmt = ret;
av_log(h->avctx, AV_LOG_VERBOSE, "Reinit context to %dx%d, "
"pix_fmt: %s\n", h->width, h->height, av_get_pix_fmt_name(h->avctx->pix_fmt));
if ((ret = h264_slice_header_init(h)) < 0) {
av_log(h->avctx, AV_LOG_ERROR,
"h264_slice_header_init() failed\n");
return ret;
}
}
return 0;
}
/* This function is called right after decoding the slice header for a first
* slice in a field (or a frame). It decides whether we are decoding a new frame
* or a second field in a pair and does the necessary setup.
*/
static int h264_field_start(H264Context *h, const H264SliceContext *sl, const H2645NAL *nal, int first_slice)
被调用
int ff_h264_queue_decode_slice(H264Context *h, const H2645NAL *nal)
被调用
指定海康摄像机的分辨率是1920*1080P，I帧间隔是25，视频帧率是25fps。查看AVFormatContext结构体中streams数组中的第一个元素codec
codec_id=AV_CODEC_ID_H264, width=1920, height=1080, coded_with=1920, coded_height=1088, pix_fmt=AV_PIX_FMT_YUVJ420P
疑问
coded_height为什么不是1080
解答
当前SPS的帧的宽 = (sps_info.pic_width_in_mbs_minus1 + 1) * 16
当前SPS的帧的高 = (sps_info.pic_height_in_map_units_minus1 + 1) * 16
通过如下代码计算得到的宽高（1888 x 1920）却不完全等同于源视频的宽高（1920 x 1080）。
根据编码规则计算到的宽高是1920x1088，而MP4中读取的是1920x1080。那么，如何FFmpeg是如何修正SPS里的宽高计算呢？
实际上，SPS还有crop系列字段，由crop表示帧是否被裁剪、crop_left、crop_right、crop_top和crop_bottom表示要裁剪的值得到正确的宽高值。
相关知识
sps和pps的结构参考：h264编码 里面的Sequence parameter set RBSP syntax
0x00000001或者0x000001是起始码，0x67是sps的开头，0x68是pps的开头。
0x42代表profile_idc,后面八位是constraint_set0_flag和reserved_zero_4bits，都设为0，0x0a是level_idc,
接着后面为图方便能用0表示的都用了。这里要注意是ue(v)表示该域是可变位，使用的指数-哥伦布编码
我的目的主要是设置正确帧高度和帧高度，所以只要填充 pic_width_in_mbs_minus1和 pic_height_in_map_units_minus1，
将它们的十六进制数写入sps_pps，
如果宽度和高度不是16的倍数，要填frame_cropping_flag
参考
http://m.blog.csdn.net/stn_lcd/article/details/74390636
http://m.blog.csdn.net/leixiaohua1020/article/details/45001033
static int h264_init_ps(H264Context *h, const H264SliceContext *sl, int first_slice)
{
const SPS *sps;
int needs_reinit = 0, must_reinit, ret;
if (first_slice) {
av_buffer_unref(&h->ps.pps_ref);
h->ps.pps = NULL;
h->ps.pps_ref = av_buffer_ref(h->ps.pps_list[sl->pps_id]);
if (!h->ps.pps_ref)
return AVERROR(ENOMEM);
h->ps.pps = (const PPS*)h->ps.pps_ref->data;
}
if (h->ps.sps != (const SPS*)h->ps.sps_list[h->ps.pps->sps_id]->data) {
av_buffer_unref(&h->ps.sps_ref);
h->ps.sps = NULL;
h->ps.sps_ref = av_buffer_ref(h->ps.sps_list[h->ps.pps->sps_id]);
if (!h->ps.sps_ref)
return AVERROR(ENOMEM);
h->ps.sps = (const SPS*)h->ps.sps_ref->data;
if (h->mb_width != h->ps.sps->mb_width ||
h->mb_height != h->ps.sps->mb_height ||
h->cur_bit_depth_luma != h->ps.sps->bit_depth_luma ||
h->cur_chroma_format_idc != h->ps.sps->chroma_format_idc
)
needs_reinit = 1;
if (h->bit_depth_luma != h->ps.sps->bit_depth_luma ||
h->chroma_format_idc != h->ps.sps->chroma_format_idc)
needs_reinit = 1;
}
sps = h->ps.sps;
must_reinit = (h->context_initialized &&
( 16*sps->mb_width != h->avctx->coded_width
|| 16*sps->mb_height != h->avctx->coded_height
|| h->cur_bit_depth_luma != sps->bit_depth_luma
|| h->cur_chroma_format_idc != sps->chroma_format_idc
|| h->mb_width != sps->mb_width
|| h->mb_height != sps->mb_height
));
if (h->avctx->pix_fmt == AV_PIX_FMT_NONE
|| (non_j_pixfmt(h->avctx->pix_fmt) != non_j_pixfmt(get_pixel_format(h, 0))))
must_reinit = 1;
if (first_slice && av_cmp_q(sps->sar, h->avctx->sample_aspect_ratio))
must_reinit = 1;
if (!h->setup_finished) {
h->avctx->profile = ff_h264_get_profile(sps);
h->avctx->level = sps->level_idc;
h->avctx->refs = sps->ref_frame_count;
h->mb_width = sps->mb_width;
h->mb_height = sps->mb_height;
h->mb_num = h->mb_width * h->mb_height;
h->mb_stride = h->mb_width + 1;
h->b_stride = h->mb_width * 4;
h->chroma_y_shift = sps->chroma_format_idc <= 1; // 400 uses yuv420p
h->width = 16 * h->mb_width;
h->height = 16 * h->mb_height;
ret = init_dimensions(h);
if (ret < 0)
return ret;
if (sps->video_signal_type_present_flag) {
h->avctx->color_range = sps->full_range > 0 ? AVCOL_RANGE_JPEG
: AVCOL_RANGE_MPEG;
if (sps->colour_description_present_flag) {
if (h->avctx->colorspace != sps->colorspace)
needs_reinit = 1;
h->avctx->color_primaries = sps->color_primaries;
h->avctx->color_trc = sps->color_trc;
h->avctx->colorspace = sps->colorspace;
}
}
}
if (!h->context_initialized || must_reinit || needs_reinit) {
int flush_changes = h->context_initialized;
h->context_initialized = 0;
if (sl != h->slice_ctx) {
av_log(h->avctx, AV_LOG_ERROR,
"changing width %d -> %d / height %d -> %d on "
"slice %d\n",
h->width, h->avctx->coded_width,
h->height, h->avctx->coded_height,
h->current_slice + 1);
return AVERROR_INVALIDDATA;
}
av_assert1(first_slice);
if (flush_changes)
ff_h264_flush_change(h);
if ((ret = get_pixel_format(h, 1)) < 0)
return ret;
h->avctx->pix_fmt = ret;
av_log(h->avctx, AV_LOG_VERBOSE, "Reinit context to %dx%d, "
"pix_fmt: %s\n", h->width, h->height, av_get_pix_fmt_name(h->avctx->pix_fmt));
if ((ret = h264_slice_header_init(h)) < 0) {
av_log(h->avctx, AV_LOG_ERROR,
"h264_slice_header_init() failed\n");
return ret;
}
}
return 0;
}
/* This function is called right after decoding the slice header for a first
* slice in a field (or a frame). It decides whether we are decoding a new frame
* or a second field in a pair and does the necessary setup.
*/
static int h264_field_start(H264Context *h, const H264SliceContext *sl, const H2645NAL *nal, int first_slice)
被调用
int ff_h264_queue_decode_slice(H264Context *h, const H2645NAL *nal)
被调用
static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size)
本文转自fengyuzaitu 51CTO博客，原文链接：http://blog.51cto.com/fengyuzaitu/2055607，如需转载请自行联系原作者
