FFMPEG结构体分析:AVFrame

本文详细分析了FFMPEG中AVFrame结构体的主要变量,包括data、linesize、width、height等,并解释了它们在视频和音频数据处理中的作用。同时,介绍了AVFrame在编码和解码过程中的应用,以及如何通过这些变量进行码流分析。

摘要生成于 C知道 ,由 DeepSeek-R1 满血版支持, 前往体验 >

FFMPEG有几个最重要的结构体,包含了解协议,解封装,解码操作,此前已经进行过分析:

FFMPEG中最关键的结构体之间的关系

在此不再详述,其中AVFrame是包含码流参数较多的结构体。本文将会详细分析一下该结构体里主要变量的含义和作用。

首先看一下结构体的定义(位于avcodec.h):

typedef struct AVFrame {
#define AV_NUM_DATA_POINTERS 8
    /**
     * pointer to the picture/channel planes.
     * This might be different from the first allocated byte
     *
     * Some decoders access areas outside 0,0 - width,height, please
     * see avcodec_align_dimensions2(). Some filters and swscale can read
     * up to 16 bytes beyond the planes, if these filters are to be used,
     * then 16 extra bytes must be allocated.
     */
    uint8_t *data[AV_NUM_DATA_POINTERS];

    /**
     * For video, size in bytes of each picture line.
     * For audio, size in bytes of each plane.
     *
     * For audio, only linesize[0] may be set. For planar audio, each channel
     * plane must be the same size.
     *
     * For video the linesizes should be multiplies(乘) of the CPUs alignment
     * preference(对齐的偏好), this is 16 or 32 for modern desktop CPUs.
     * Some code requires such alignment other code can be slower without
     * correct alignment, for yet other it makes no difference.
     *
     * @note The linesize may be larger than the size of usable data -- there
     * may be extra padding present for performance reasons.
     */
    int linesize[AV_NUM_DATA_POINTERS];

    /**
     * pointers to the data planes/channels.
     *
     * For video, this should simply point to data[].
     *
     * For planar(平面) audio, each channel has a separate(单独的) data pointer, and
     * linesize[0] contains the size of each channel buffer.
     * For packed audio, there is just one data pointer, and linesize[0]
     * contains the total size of the buffer for all channels.
     *
     * Note: Both data and extended_data should always be set in a valid(有效的) frame,
     * but for planar audio with more channels that can fit(适应) in data,
     * extended_data must be used in order to access all channels.
     */
    uint8_t **extended_data;

    /**
     * width and height of the video frame
     */
    int width, height;

    /**
     * number of audio samples (per channel) described by this frame
     */
    int nb_samples;

    /**
     * format of the frame, -1 if unknown or unset
     * Values correspond(符合) to enum AVPixelFormat for video frames,
     * enum AVSampleFormat for audio)
     */
    int format; //AV_PIX_FMT_YUV420P  AV_SAMPLE_FMT_NB

    /**
     * 1 -> keyframe, 0-> not
     */
    int key_frame;

    /**
     * Picture type of the frame.(帧类型:i p b)
     */
    enum AVPictureType pict_type;

#if FF_API_AVFRAME_LAVC
    attribute_deprecated
    uint8_t *base[AV_NUM_DATA_POINTERS];
#endif

    /**
     * Sample aspect ratio for the video frame, 0/1 if unknown/unspecified.
     * 对于视频帧采样宽高比,0 / 1如果未知/未指定。
     */
    AVRational sample_aspect_ratio;

    /**
     * 显示时间戳
     * Presentation timestamp in time_base units (time when frame should be shown to user).
     */
    int64_t pts;

    /**
     * 解码时间戳
     * PTS copied from the AVPacket that was decoded to produce this frame.
     */
    int64_t pkt_pts;

    /**
     * DTS copied from the AVPacket that triggered returning this frame. (if frame threading isnt used)
     * This is also the Presentation time of this AVFrame calculated from
     * only AVPacket.dts values without pts values.
     */
    int64_t pkt_dts;

    /**
     * picture number in bitstream order
     */
    int coded_picture_number;
    /**
     * picture number in display order
     */
    int display_picture_number;

    /**
     * quality 质量(between 1 (good) and FF_LAMBDA_MAX (bad))
     */
    int quality;

#if FF_API_AVFRAME_LAVC
    attribute_deprecated
    int reference;

    /**
     * QP table QP表
     */
    attribute_deprecated
    int8_t *qscale_table;
    /**
     * QP store stride (步幅)
     */
    attribute_deprecated
    int qstride;

    attribute_deprecated
    int qscale_type;

    /**
     * 跳过宏块表
     * mbskip_table[mb]>=1 if MB didn't change
     * stride= mb_width = (width+15)>>4
     */
    attribute_deprecated
    uint8_t *mbskip_table;

    /**
     * 运动矢量表
     * motion vector table
     * @code
     * example:
     * int mv_sample_log2= 4 - motion_subsample_log2;
     * int mb_width= (width+15)>>4;
     * int mv_stride= (mb_width << mv_sample_log2) + 1;
     * motion_val[direction][x + y*mv_stride][0->mv_x, 1->mv_y];
     * @endcode
     */
    attribute_deprecated
    int16_t (*motion_val[2])[2];

    /**
     * 宏块类型表
     * macroblock type table
     * mb_type_base + mb_width + 2
     */
    attribute_deprecated
    uint32_t *mb_type;

    /**
     * DCT系数
     * DCT coefficients
     */
    attribute_deprecated
    short *dct_coeff;

    /**
     * 参考帧列表
     * motion reference frame index
     * the order in which these are stored can depend on the codec.
     */
    attribute_deprecated
    int8_t *ref_index[2];
#endif

    /**
     * for some private data of the user
     */
    void *opaque;

    /**
     * error
     */
    uint64_t error[AV_NUM_DATA_POINTERS];

#if FF_API_AVFRAME_LAVC
    attribute_deprecated
    int type;
#endif

    /**
     * When decoding, this signals how much the picture must be delayed.
     * 解码时,该信号多少图片必须延迟。
     * extra_delay = repeat_pict / (2*fps)
     */
    int repeat_pict;

    /**
     * The content of the picture is interlaced(隔行扫描).
     */
    int interlaced_frame;

    /**
     * If the content is interlaced, is top field displayed first.
     * 如果是隔行扫描,是否是顶场先被显示
     */
    int top_field_first;

    /**
     * Tell user application that palette(调色板) has changed from previous frame.
     */
    int palette_has_changed;

#if FF_API_AVFRAME_LAVC
    attribute_deprecated
    int buffer_hints;

    /**
     * Pan scan.  pan 扫描
     */
    attribute_deprecated
    struct AVPanScan *pan_scan;
#endif

    /**
     * reordered opaque 64bit (generally an integer or a double precision float
     * PTS but can be anything).
     * The user sets AVCodecContext.reordered_opaque to represent the input at
     * that time,
     * the decoder reorders values as needed and sets AVFrame.reordered_opaque
     * to exactly one of the values provided by the user through AVCodecContext.reordered_opaque
     * @deprecated in favor of pkt_pts
     * 重新排序(通常是64位的整型或不透明双精密浮点但可以做)。
     * 不透明的用户集AVCodecContext.reordered_opaque代表输入到解码器是在时间的重新排序的值,
     * 需要对项集avframe.reordered _不透明值,一个由用户提供的avcodeccontext.reordered暗通”请取消_(PTS中_袋
     */
    int64_t reordered_opaque;

#if FF_API_AVFRAME_LAVC
    /**
     * @deprecated this field is unused
     */
    attribute_deprecated void *hwaccel_picture_private;

    attribute_deprecated
    struct AVCodecContext *owner;
    attribute_deprecated
    void *thread_opaque;

    /**
     * log2 of the size of the block which a single vector in motion_val represents:
     * (4->16x16, 3->8x8, 2-> 4x4, 1-> 2x2)
     */
    attribute_deprecated
    uint8_t motion_subsample_log2;
#endif

    /**
     * Sample rate of the audio data.
     */
    int sample_rate;

    /**
     * Channel layout of the audio data.
     */
    uint64_t channel_layout;

    /**
     * AVBuffer references backing the data for this frame. If all elements of
     * this array are NULL, then this frame is not reference counted.
     *
     * There may be at most one AVBuffer per data plane, so for video this array
     * always contains all the references. For planar audio with more than
     * AV_NUM_DATA_POINTERS channels, there may be more buffers than can fit in
     * this array. Then the extra AVBufferRef pointers are stored in the
     * extended_buf array.
     */
    AVBufferRef *buf[AV_NUM_DATA_POINTERS];

    /**
     * For planar audio which requires more than AV_NUM_DATA_POINTERS
     * AVBufferRef pointers, this array will hold all the references which
     * cannot fit into AVFrame.buf.
     *
     * Note that this is different from AVFrame.extended_data, which always
     * contains all the pointers. This array only contains the extra pointers,
     * which cannot fit into AVFrame.buf.
     *
     * This array is always allocated using av_malloc() by whoever constructs
     * the frame. It is freed in av_frame_unref().
     */
    AVBufferRef **extended_buf;
    /**
     * Number of elements in extended_buf.
     */
    int        nb_extended_buf;

    AVFrameSideData **side_data;
    int            nb_side_data;

    /**
     * frame timestamp estimated using various heuristics, in stream time base
     * Code outside libavcodec should access this field using:
     * av_frame_get_best_effort_timestamp(frame)
     * - encoding: unused
     * - decoding: set by libavcodec, read by user.
     */
    int64_t best_effort_timestamp;

    /**
     * reordered pos from the last AVPacket that has been input into the decoder
     * Code outside libavcodec should access this field using:
     * av_frame_get_pkt_pos(frame)
     * - encoding: unused
     * - decoding: Read by user.
     */
    int64_t pkt_pos;

    /**
     * duration of the corresponding packet, expressed in
     * AVStream->time_base units, 0 if unknown.
     * Code outside libavcodec should access this field using:
     * av_frame_get_pkt_duration(frame)
     * - encoding: unused
     * - decoding: Read by user.
     */
    int64_t pkt_duration;

    /**
     * metadata.
     * Code outside libavcodec should access this field using:
     * av_frame_get_metadata(frame)
     * - encoding: Set by user.
     * - decoding: Set by libavcodec.
     */
    AVDictionary *metadata;

    /**
     * decode error flags of the frame, set to a combination of
     * FF_DECODE_ERROR_xxx flags if the decoder produced a frame, but there
     * were errors during the decoding.
     * Code outside libavcodec should access this field using:
     * av_frame_get_decode_error_flags(frame)
     * - encoding: unused
     * - decoding: set by libavcodec, read by user.
     */
    int decode_error_flags;
#define FF_DECODE_ERROR_INVALID_BITSTREAM   1
#define FF_DECODE_ERROR_MISSING_REFERENCE   2

    /**
     * number of audio channels, only used for audio.
     * Code outside libavcodec should access this field using:
     * av_frame_get_channels(frame)
     * - encoding: unused
     * - decoding: Read by user.
     */
    int channels;

    /**
     * size of the corresponding packet containing the compressed
     * frame. It must be accessed using av_frame_get_pkt_size() and
     * av_frame_set_pkt_size().
     * It is set to a negative value if unknown.
     * - encoding: unused
     * - decoding: set by libavcodec, read by user.
     */
    int pkt_size;

    /**
     * YUV colorspace type.
     * It must be accessed using av_frame_get_colorspace() and
     * av_frame_set_colorspace().
     * - encoding: Set by user
     * - decoding: Set by libavcodec
     */
    enum AVColorSpace colorspace;

    /**
     * MPEG vs JPEG YUV range.
     * It must be accessed using av_frame_get_color_range() and
     * av_frame_set_color_range().
     * - encoding: Set by user
     * - decoding: Set by libavcodec
     */
    enum AVColorRange color_range;


    /**
     * Not to be accessed directly from outside libavutil
     */
    AVBufferRef *qp_table_buf;
} AVFrame;


<span style="font-family: Arial; font-size: 14px; line-height: 26px; background-color: rgb(255, 255, 255);">AVFrame结构体一般用于存储原始数据(即非压缩数据,例如对视频来说是YUV,RGB,对音频来说是PCM),此外还包含了一些相关的信息。比如说,解码的时候存储了宏块类型表,QP表,运动矢量表等数据。编码的时候也存储了相关的数据。因此在使用FFMPEG进行码流分析的时候,AVFrame是一个很重要的结构体。</span>

下面看几个主要变量的作用(在这里考虑解码的情况)

uint8_t *data[AV_NUM_DATA_POINTERS]:解码后原始数据(对视频来说是YUV,RGB,对音频来说是PCM)
int linesize[AV_NUM_DATA_POINTERS]:data的大小
int width, height:视频帧宽和高(1920x1080,1280x720...)
int nb_samples:音频的一个AVFrame中可能包含多个音频帧,在此标记包含了几个
int format:解码后原始数据类型(YUV420,YUV422,RGB24...)
int key_frame:是否是关键帧
enum AVPictureType pict_type:帧类型(I,B,P...)
AVRational sample_aspect_ratio:宽高比(16:9,4:3...)
int64_t pts:显示时间戳
int coded_picture_number:编码帧序号
int display_picture_number:显示帧序号
int8_t *qscale_table:QP表
uint8_t *mbskip_table:跳过宏块表
int16_t (*motion_val[2])[2]:运动矢量表
uint32_t *mb_type:宏块类型表
short *dct_coeff:DCT系数,这个没有提取过
int8_t *ref_index[2]:运动估计参考帧列表(貌似H.264这种比较新的标准才会涉及到多参考帧)
int interlaced_frame:是否是隔行扫描
uint8_t motion_subsample_log2:一个宏块中的运动矢量采样个数,取log的
其他的变量不再一一列举,源代码中都有详细的说明。在这里重点分析一下几个需要一定的理解的变量:

1.data[]

对于packed格式的数据(例如RGB24),会存到data[0]里面。

对于planar格式的数据(例如YUV420P),则会分开成data[0],data[1],data[2]...(YUV420P中data[0]存Y,data[1]存U,data[2]存V)

具体参见:FFMPEG 实现 YUV,RGB各种图像原始数据之间的转换(swscale)

2.pict_type

包含以下类型:

enum AVPictureType {
    AV_PICTURE_TYPE_NONE = 0, ///< Undefined
    AV_PICTURE_TYPE_I,     ///< Intra
    AV_PICTURE_TYPE_P,     ///< Predicted
    AV_PICTURE_TYPE_B,     ///< Bi-dir predicted
    AV_PICTURE_TYPE_S,     ///< S(GMC)-VOP MPEG4
    AV_PICTURE_TYPE_SI,    ///< Switching Intra
    AV_PICTURE_TYPE_SP,    ///< Switching Predicted
    AV_PICTURE_TYPE_BI,    ///< BI type
};
3.sample_aspect_ratio

宽高比是一个分数,FFMPEG中用AVRational表达分数:

/**
 * rational number numerator/denominator
 */
typedef struct AVRational{
    int num; ///< numerator
    int den; ///< denominator
} AVRational;
4.qscale_table

QP表指向一块内存,里面存储的是每个宏块的QP值。宏块的标号是从左往右,一行一行的来的。每个宏块对应1个QP。

qscale_table[0]就是第1行第1列宏块的QP值;qscale_table[1]就是第1行第2列宏块的QP值;qscale_table[2]就是第1行第3列宏块的QP值。以此类推...

宏块的个数用下式计算:

注:宏块大小是16x16的。

每行宏块数:

int mb_stride = pCodecCtx->width/16+1
宏块的总数:

int mb_sum = ((pCodecCtx->height+15)>>4)*(pCodecCtx->width/16+1)

5.motion_subsample_log2

1个运动矢量所能代表的画面大小(用宽或者高表示,单位是像素),注意,这里取了log2。

代码注释中给出以下数据:

4->16x16, 3->8x8, 2-> 4x4, 1-> 2x2

即1个运动矢量代表16x16的画面的时候,该值取4;1个运动矢量代表8x8的画面的时候,该值取3...以此类推

6.motion_val

运动矢量表存储了一帧视频中的所有运动矢量。

该值的存储方式比较特别:

int16_t (*motion_val[2])[2];
为了弄清楚该值究竟是怎么存的,花了我好一阵子功夫...

注释中给了一段代码:

int mv_sample_log2= 4 - motion_subsample_log2;
int mb_width= (width+15)>>4;
int mv_stride= (mb_width << mv_sample_log2) + 1;
motion_val[direction][x + y*mv_stride][0->mv_x, 1->mv_y];

大概知道了该数据的结构:

1.首先分为两个列表L0和L1

2.每个列表(L0或L1)存储了一系列的MV(每个MV对应一个画面,大小由motion_subsample_log2决定)

3.每个MV分为横坐标和纵坐标(x,y)

注意,在FFMPEG中MV和MB在存储的结构上是没有什么关联的,第1个MV是屏幕上左上角画面的MV(画面的大小取决于motion_subsample_log2),第2个MV是屏幕上第1行第2列的画面的MV,以此类推。因此在一个宏块(16x16)的运动矢量很有可能如下图所示(line代表一行运动矢量的个数):

//例如8x8划分的运动矢量与宏块的关系:
				//-------------------------
				//|          |            |
				//|mv[x]     |mv[x+1]     |
				//-------------------------
				//|          |	          |
				//|mv[x+line]|mv[x+line+1]|
				//-------------------------

7.mb_type

宏块类型表存储了一帧视频中的所有宏块的类型。其存储方式和QP表差不多。只不过其是uint32类型的,而QP表是uint8类型的。每个宏块对应一个宏块类型变量。

宏块类型如下定义所示:

//The following defines may change, don't expect compatibility if you use them.
#define MB_TYPE_INTRA4x4   0x0001
#define MB_TYPE_INTRA16x16 0x0002 //FIXME H.264-specific
#define MB_TYPE_INTRA_PCM  0x0004 //FIXME H.264-specific
#define MB_TYPE_16x16      0x0008
#define MB_TYPE_16x8       0x0010
#define MB_TYPE_8x16       0x0020
#define MB_TYPE_8x8        0x0040
#define MB_TYPE_INTERLACED 0x0080
#define MB_TYPE_DIRECT2    0x0100 //FIXME
#define MB_TYPE_ACPRED     0x0200
#define MB_TYPE_GMC        0x0400
#define MB_TYPE_SKIP       0x0800
#define MB_TYPE_P0L0       0x1000
#define MB_TYPE_P1L0       0x2000
#define MB_TYPE_P0L1       0x4000
#define MB_TYPE_P1L1       0x8000
#define MB_TYPE_L0         (MB_TYPE_P0L0 | MB_TYPE_P1L0)
#define MB_TYPE_L1         (MB_TYPE_P0L1 | MB_TYPE_P1L1)
#define MB_TYPE_L0L1       (MB_TYPE_L0   | MB_TYPE_L1)
#define MB_TYPE_QUANT      0x00010000
#define MB_TYPE_CBP        0x00020000
//Note bits 24-31 are reserved for codec specific use (h264 ref0, mpeg1 0mv, ...)

一个宏块如果包含上述定义中的一种或两种类型,则其对应的宏块变量的对应位会被置1。
注:一个宏块可以包含好几种类型,但是有些类型是不能重复包含的,比如说一个宏块不可能既是16x16又是8x8。


8.ref_index

运动估计参考帧列表存储了一帧视频中所有宏块的参考帧索引。这个列表其实在比较早的压缩编码标准中是没有什么用的。只有像H.264这样的编码标准才有多参考帧的概念。但是这个字段目前我还没有研究透。只是知道每个宏块包含有4个该值,该值反映的是参考帧的索引。以后有机会再进行细研究吧。


在这里展示一下自己做的码流分析软件的运行结果。将上文介绍的几个列表图像化显示了出来(在这里是使用MFC的绘图函数画出来的)

视频帧:


QP参数提取的结果:


美化过的(加上了颜色):


宏块类型参数提取的结果:


美化过的(加上了颜色,更清晰一些,s代表skip宏块):



运动矢量参数提取的结果(在这里是List0):


运动估计参考帧参数提取的结果:

来源:http://blog.youkuaiyun.com/leixiaohua1020/article/details/14214577









评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值