2 * This file is part of FFmpeg.
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 * reference-counted frame API
25 #ifndef AVUTIL_FRAME_H
26 #define AVUTIL_FRAME_H
35 #include "samplefmt.h"
41 * @defgroup lavu_frame AVFrame
45 * AVFrame is an abstraction for reference-counted raw multimedia data.
48 enum AVFrameSideDataType {
50 * The data is the AVPanScan struct defined in libavcodec.
52 AV_FRAME_DATA_PANSCAN,
54 * ATSC A53 Part 4 Closed Captions.
55 * A53 CC bitstream is stored as uint8_t in AVFrameSideData.data.
56 * The number of bytes of CC data is AVFrameSideData.size.
60 * Stereoscopic 3d metadata.
61 * The data is the AVStereo3D struct defined in libavutil/stereo3d.h.
63 AV_FRAME_DATA_STEREO3D,
65 * The data is the AVMatrixEncoding enum defined in libavutil/channel_layout.h.
67 AV_FRAME_DATA_MATRIXENCODING,
69 * Metadata relevant to a downmix procedure.
70 * The data is the AVDownmixInfo struct defined in libavutil/downmix_info.h.
72 AV_FRAME_DATA_DOWNMIX_INFO,
74 * ReplayGain information in the form of the AVReplayGain struct.
76 AV_FRAME_DATA_REPLAYGAIN,
78 * This side data contains a 3x3 transformation matrix describing an affine
79 * transformation that needs to be applied to the frame for correct
82 * See libavutil/display.h for a detailed description of the data.
84 AV_FRAME_DATA_DISPLAYMATRIX,
86 * Active Format Description data consisting of a single byte as specified
87 * in ETSI TS 101 154 using AVActiveFormatDescription enum.
91 * Motion vectors exported by some codecs (on demand through the export_mvs
92 * flag set in the libavcodec AVCodecContext flags2 option).
93 * The data is the AVMotionVector struct defined in
94 * libavutil/motion_vector.h.
96 AV_FRAME_DATA_MOTION_VECTORS,
98 * Recommmends skipping the specified number of samples. This is exported
99 * only if the "skip_manual" AVOption is set in libavcodec.
100 * This has the same format as AV_PKT_DATA_SKIP_SAMPLES.
102 * u32le number of samples to skip from start of this packet
103 * u32le number of samples to skip from end of this packet
104 * u8 reason for start skip
105 * u8 reason for end skip (0=padding silence, 1=convergence)
108 AV_FRAME_DATA_SKIP_SAMPLES,
110 * This side data must be associated with an audio frame and corresponds to
111 * enum AVAudioServiceType defined in avcodec.h.
113 AV_FRAME_DATA_AUDIO_SERVICE_TYPE,
115 * Mastering display metadata associated with a video frame. The payload is
116 * an AVMasteringDisplayMetadata type and contains information about the
117 * mastering display color volume.
119 AV_FRAME_DATA_MASTERING_DISPLAY_METADATA,
121 * The GOP timecode in 25 bit timecode format. Data format is 64-bit integer.
122 * This is set on the first frame of a GOP that has a temporal reference of 0.
124 AV_FRAME_DATA_GOP_TIMECODE,
127 * The data represents the AVSphericalMapping structure defined in
128 * libavutil/spherical.h.
130 AV_FRAME_DATA_SPHERICAL,
133 * Content light level (based on CTA-861.3). This payload contains data in
134 * the form of the AVContentLightMetadata struct.
136 AV_FRAME_DATA_CONTENT_LIGHT_LEVEL,
139 * The data contains an ICC profile as an opaque octet buffer following the
140 * format described by ISO 15076-1 with an optional name defined in the
141 * metadata key entry "name".
143 AV_FRAME_DATA_ICC_PROFILE,
146 enum AVActiveFormatDescription {
151 AV_AFD_4_3_SP_14_9 = 13,
152 AV_AFD_16_9_SP_14_9 = 14,
158 * Structure to hold side data for an AVFrame.
160 * sizeof(AVFrameSideData) is not a part of the public ABI, so new fields may be added
161 * to the end with a minor bump.
163 typedef struct AVFrameSideData {
164 enum AVFrameSideDataType type;
167 AVDictionary *metadata;
172 * This structure describes decoded (raw) audio or video data.
174 * AVFrame must be allocated using av_frame_alloc(). Note that this only
175 * allocates the AVFrame itself, the buffers for the data must be managed
176 * through other means (see below).
177 * AVFrame must be freed with av_frame_free().
179 * AVFrame is typically allocated once and then reused multiple times to hold
180 * different data (e.g. a single AVFrame to hold frames received from a
181 * decoder). In such a case, av_frame_unref() will free any references held by
182 * the frame and reset it to its original clean state before it
185 * The data described by an AVFrame is usually reference counted through the
186 * AVBuffer API. The underlying buffer references are stored in AVFrame.buf /
187 * AVFrame.extended_buf. An AVFrame is considered to be reference counted if at
188 * least one reference is set, i.e. if AVFrame.buf[0] != NULL. In such a case,
189 * every single data plane must be contained in one of the buffers in
190 * AVFrame.buf or AVFrame.extended_buf.
191 * There may be a single buffer for all the data, or one separate buffer for
192 * each plane, or anything in between.
194 * sizeof(AVFrame) is not a part of the public ABI, so new fields may be added
195 * to the end with a minor bump.
197 * Fields can be accessed through AVOptions, the name string used, matches the
198 * C structure field name for fields accessible through AVOptions. The AVClass
199 * for AVFrame can be obtained from avcodec_get_frame_class()
201 typedef struct AVFrame {
202 #define AV_NUM_DATA_POINTERS 8
204 * pointer to the picture/channel planes.
205 * This might be different from the first allocated byte
207 * Some decoders access areas outside 0,0 - width,height, please
208 * see avcodec_align_dimensions2(). Some filters and swscale can read
209 * up to 16 bytes beyond the planes, if these filters are to be used,
210 * then 16 extra bytes must be allocated.
212 * NOTE: Except for hwaccel formats, pointers not needed by the format
213 * MUST be set to NULL.
215 uint8_t *data[AV_NUM_DATA_POINTERS];
218 * For video, size in bytes of each picture line.
219 * For audio, size in bytes of each plane.
221 * For audio, only linesize[0] may be set. For planar audio, each channel
222 * plane must be the same size.
224 * For video the linesizes should be multiples of the CPUs alignment
225 * preference, this is 16 or 32 for modern desktop CPUs.
226 * Some code requires such alignment other code can be slower without
227 * correct alignment, for yet other it makes no difference.
229 * @note The linesize may be larger than the size of usable data -- there
230 * may be extra padding present for performance reasons.
232 int linesize[AV_NUM_DATA_POINTERS];
235 * pointers to the data planes/channels.
237 * For video, this should simply point to data[].
239 * For planar audio, each channel has a separate data pointer, and
240 * linesize[0] contains the size of each channel buffer.
241 * For packed audio, there is just one data pointer, and linesize[0]
242 * contains the total size of the buffer for all channels.
244 * Note: Both data and extended_data should always be set in a valid frame,
245 * but for planar audio with more channels that can fit in data,
246 * extended_data must be used in order to access all channels.
248 uint8_t **extended_data;
251 * @name Video dimensions
252 * Video frames only. The coded dimensions (in pixels) of the video frame,
253 * i.e. the size of the rectangle that contains some well-defined values.
255 * @note The part of the frame intended for display/presentation is further
256 * restricted by the @ref cropping "Cropping rectangle".
265 * number of audio samples (per channel) described by this frame
270 * format of the frame, -1 if unknown or unset
271 * Values correspond to enum AVPixelFormat for video frames,
272 * enum AVSampleFormat for audio)
277 * 1 -> keyframe, 0-> not
282 * Picture type of the frame.
284 enum AVPictureType pict_type;
287 * Sample aspect ratio for the video frame, 0/1 if unknown/unspecified.
289 AVRational sample_aspect_ratio;
292 * Presentation timestamp in time_base units (time when frame should be shown to user).
298 * PTS copied from the AVPacket that was decoded to produce this frame.
299 * @deprecated use the pts field instead
306 * DTS copied from the AVPacket that triggered returning this frame. (if frame threading isn't used)
307 * This is also the Presentation time of this AVFrame calculated from
308 * only AVPacket.dts values without pts values.
313 * picture number in bitstream order
315 int coded_picture_number;
317 * picture number in display order
319 int display_picture_number;
322 * quality (between 1 (good) and FF_LAMBDA_MAX (bad))
327 * for some private data of the user
331 #if FF_API_ERROR_FRAME
336 uint64_t error[AV_NUM_DATA_POINTERS];
340 * When decoding, this signals how much the picture must be delayed.
341 * extra_delay = repeat_pict / (2*fps)
346 * The content of the picture is interlaced.
348 int interlaced_frame;
351 * If the content is interlaced, is top field displayed first.
356 * Tell user application that palette has changed from previous frame.
358 int palette_has_changed;
361 * reordered opaque 64 bits (generally an integer or a double precision float
362 * PTS but can be anything).
363 * The user sets AVCodecContext.reordered_opaque to represent the input at
365 * the decoder reorders values as needed and sets AVFrame.reordered_opaque
366 * to exactly one of the values provided by the user through AVCodecContext.reordered_opaque
367 * @deprecated in favor of pkt_pts
369 int64_t reordered_opaque;
372 * Sample rate of the audio data.
377 * Channel layout of the audio data.
379 uint64_t channel_layout;
382 * AVBuffer references backing the data for this frame. If all elements of
383 * this array are NULL, then this frame is not reference counted. This array
384 * must be filled contiguously -- if buf[i] is non-NULL then buf[j] must
385 * also be non-NULL for all j < i.
387 * There may be at most one AVBuffer per data plane, so for video this array
388 * always contains all the references. For planar audio with more than
389 * AV_NUM_DATA_POINTERS channels, there may be more buffers than can fit in
390 * this array. Then the extra AVBufferRef pointers are stored in the
391 * extended_buf array.
393 AVBufferRef *buf[AV_NUM_DATA_POINTERS];
396 * For planar audio which requires more than AV_NUM_DATA_POINTERS
397 * AVBufferRef pointers, this array will hold all the references which
398 * cannot fit into AVFrame.buf.
400 * Note that this is different from AVFrame.extended_data, which always
401 * contains all the pointers. This array only contains the extra pointers,
402 * which cannot fit into AVFrame.buf.
404 * This array is always allocated using av_malloc() by whoever constructs
405 * the frame. It is freed in av_frame_unref().
407 AVBufferRef **extended_buf;
409 * Number of elements in extended_buf.
413 AVFrameSideData **side_data;
417 * @defgroup lavu_frame_flags AV_FRAME_FLAGS
418 * @ingroup lavu_frame
419 * Flags describing additional frame properties.
425 * The frame data may be corrupted, e.g. due to decoding errors.
427 #define AV_FRAME_FLAG_CORRUPT (1 << 0)
429 * A flag to mark the frames which need to be decoded, but shouldn't be output.
431 #define AV_FRAME_FLAG_DISCARD (1 << 2)
437 * Frame flags, a combination of @ref lavu_frame_flags
442 * MPEG vs JPEG YUV range.
443 * - encoding: Set by user
444 * - decoding: Set by libavcodec
446 enum AVColorRange color_range;
448 enum AVColorPrimaries color_primaries;
450 enum AVColorTransferCharacteristic color_trc;
453 * YUV colorspace type.
454 * - encoding: Set by user
455 * - decoding: Set by libavcodec
457 enum AVColorSpace colorspace;
459 enum AVChromaLocation chroma_location;
462 * frame timestamp estimated using various heuristics, in stream time base
464 * - decoding: set by libavcodec, read by user.
466 int64_t best_effort_timestamp;
469 * reordered pos from the last AVPacket that has been input into the decoder
471 * - decoding: Read by user.
476 * duration of the corresponding packet, expressed in
477 * AVStream->time_base units, 0 if unknown.
479 * - decoding: Read by user.
481 int64_t pkt_duration;
485 * - encoding: Set by user.
486 * - decoding: Set by libavcodec.
488 AVDictionary *metadata;
491 * decode error flags of the frame, set to a combination of
492 * FF_DECODE_ERROR_xxx flags if the decoder produced a frame, but there
493 * were errors during the decoding.
495 * - decoding: set by libavcodec, read by user.
497 int decode_error_flags;
498 #define FF_DECODE_ERROR_INVALID_BITSTREAM 1
499 #define FF_DECODE_ERROR_MISSING_REFERENCE 2
502 * number of audio channels, only used for audio.
504 * - decoding: Read by user.
509 * size of the corresponding packet containing the compressed
511 * It is set to a negative value if unknown.
513 * - decoding: set by libavcodec, read by user.
522 int8_t *qscale_table;
532 AVBufferRef *qp_table_buf;
535 * For hwaccel-format frames, this should be a reference to the
536 * AVHWFramesContext describing the frame.
538 AVBufferRef *hw_frames_ctx;
541 * AVBufferRef for free use by the API user. FFmpeg will never check the
542 * contents of the buffer ref. FFmpeg calls av_buffer_unref() on it when
543 * the frame is unreferenced. av_frame_copy_props() calls create a new
544 * reference with av_buffer_ref() for the target frame's opaque_ref field.
546 * This is unrelated to the opaque field, although it serves a similar
549 AVBufferRef *opaque_ref;
554 * Video frames only. The number of pixels to discard from the the
555 * top/bottom/left/right border of the frame to obtain the sub-rectangle of
556 * the frame intended for presentation.
569 * Accessors for some AVFrame fields. These used to be provided for ABI
570 * compatibility, and do not need to be used anymore.
572 int64_t av_frame_get_best_effort_timestamp(const AVFrame *frame);
573 void av_frame_set_best_effort_timestamp(AVFrame *frame, int64_t val);
574 int64_t av_frame_get_pkt_duration (const AVFrame *frame);
575 void av_frame_set_pkt_duration (AVFrame *frame, int64_t val);
576 int64_t av_frame_get_pkt_pos (const AVFrame *frame);
577 void av_frame_set_pkt_pos (AVFrame *frame, int64_t val);
578 int64_t av_frame_get_channel_layout (const AVFrame *frame);
579 void av_frame_set_channel_layout (AVFrame *frame, int64_t val);
580 int av_frame_get_channels (const AVFrame *frame);
581 void av_frame_set_channels (AVFrame *frame, int val);
582 int av_frame_get_sample_rate (const AVFrame *frame);
583 void av_frame_set_sample_rate (AVFrame *frame, int val);
584 AVDictionary *av_frame_get_metadata (const AVFrame *frame);
585 void av_frame_set_metadata (AVFrame *frame, AVDictionary *val);
586 int av_frame_get_decode_error_flags (const AVFrame *frame);
587 void av_frame_set_decode_error_flags (AVFrame *frame, int val);
588 int av_frame_get_pkt_size(const AVFrame *frame);
589 void av_frame_set_pkt_size(AVFrame *frame, int val);
590 AVDictionary **avpriv_frame_get_metadatap(AVFrame *frame);
592 int8_t *av_frame_get_qp_table(AVFrame *f, int *stride, int *type);
593 int av_frame_set_qp_table(AVFrame *f, AVBufferRef *buf, int stride, int type);
595 enum AVColorSpace av_frame_get_colorspace(const AVFrame *frame);
596 void av_frame_set_colorspace(AVFrame *frame, enum AVColorSpace val);
597 enum AVColorRange av_frame_get_color_range(const AVFrame *frame);
598 void av_frame_set_color_range(AVFrame *frame, enum AVColorRange val);
601 * Get the name of a colorspace.
602 * @return a static string identifying the colorspace; can be NULL.
604 const char *av_get_colorspace_name(enum AVColorSpace val);
607 * Allocate an AVFrame and set its fields to default values. The resulting
608 * struct must be freed using av_frame_free().
610 * @return An AVFrame filled with default values or NULL on failure.
612 * @note this only allocates the AVFrame itself, not the data buffers. Those
613 * must be allocated through other means, e.g. with av_frame_get_buffer() or
616 AVFrame *av_frame_alloc(void);
619 * Free the frame and any dynamically allocated objects in it,
620 * e.g. extended_data. If the frame is reference counted, it will be
621 * unreferenced first.
623 * @param frame frame to be freed. The pointer will be set to NULL.
625 void av_frame_free(AVFrame **frame);
628 * Set up a new reference to the data described by the source frame.
630 * Copy frame properties from src to dst and create a new reference for each
631 * AVBufferRef from src.
633 * If src is not reference counted, new buffers are allocated and the data is
636 * @warning: dst MUST have been either unreferenced with av_frame_unref(dst),
637 * or newly allocated with av_frame_alloc() before calling this
638 * function, or undefined behavior will occur.
640 * @return 0 on success, a negative AVERROR on error
642 int av_frame_ref(AVFrame *dst, const AVFrame *src);
645 * Create a new frame that references the same data as src.
647 * This is a shortcut for av_frame_alloc()+av_frame_ref().
649 * @return newly created AVFrame on success, NULL on error.
651 AVFrame *av_frame_clone(const AVFrame *src);
654 * Unreference all the buffers referenced by frame and reset the frame fields.
656 void av_frame_unref(AVFrame *frame);
659 * Move everything contained in src to dst and reset src.
661 * @warning: dst is not unreferenced, but directly overwritten without reading
662 * or deallocating its contents. Call av_frame_unref(dst) manually
663 * before calling this function to ensure that no memory is leaked.
665 void av_frame_move_ref(AVFrame *dst, AVFrame *src);
668 * Allocate new buffer(s) for audio or video data.
670 * The following fields must be set on frame before calling this function:
671 * - format (pixel format for video, sample format for audio)
672 * - width and height for video
673 * - nb_samples and channel_layout for audio
675 * This function will fill AVFrame.data and AVFrame.buf arrays and, if
676 * necessary, allocate and fill AVFrame.extended_data and AVFrame.extended_buf.
677 * For planar formats, one buffer will be allocated for each plane.
679 * @warning: if frame already has been allocated, calling this function will
680 * leak memory. In addition, undefined behavior can occur in certain
683 * @param frame frame in which to store the new buffers.
684 * @param align Required buffer size alignment. If equal to 0, alignment will be
685 * chosen automatically for the current CPU. It is highly
686 * recommended to pass 0 here unless you know what you are doing.
688 * @return 0 on success, a negative AVERROR on error.
690 int av_frame_get_buffer(AVFrame *frame, int align);
693 * Check if the frame data is writable.
695 * @return A positive value if the frame data is writable (which is true if and
696 * only if each of the underlying buffers has only one reference, namely the one
697 * stored in this frame). Return 0 otherwise.
699 * If 1 is returned the answer is valid until av_buffer_ref() is called on any
700 * of the underlying AVBufferRefs (e.g. through av_frame_ref() or directly).
702 * @see av_frame_make_writable(), av_buffer_is_writable()
704 int av_frame_is_writable(AVFrame *frame);
707 * Ensure that the frame data is writable, avoiding data copy if possible.
709 * Do nothing if the frame is writable, allocate new buffers and copy the data
712 * @return 0 on success, a negative AVERROR on error.
714 * @see av_frame_is_writable(), av_buffer_is_writable(),
715 * av_buffer_make_writable()
717 int av_frame_make_writable(AVFrame *frame);
720 * Copy the frame data from src to dst.
722 * This function does not allocate anything, dst must be already initialized and
723 * allocated with the same parameters as src.
725 * This function only copies the frame data (i.e. the contents of the data /
726 * extended data arrays), not any other properties.
728 * @return >= 0 on success, a negative AVERROR on error.
730 int av_frame_copy(AVFrame *dst, const AVFrame *src);
733 * Copy only "metadata" fields from src to dst.
735 * Metadata for the purpose of this function are those fields that do not affect
736 * the data layout in the buffers. E.g. pts, sample rate (for audio) or sample
737 * aspect ratio (for video), but not width/height or channel layout.
738 * Side data is also copied.
740 int av_frame_copy_props(AVFrame *dst, const AVFrame *src);
743 * Get the buffer reference a given data plane is stored in.
745 * @param plane index of the data plane of interest in frame->extended_data.
747 * @return the buffer reference that contains the plane or NULL if the input
748 * frame is not valid.
750 AVBufferRef *av_frame_get_plane_buffer(AVFrame *frame, int plane);
753 * Add a new side data to a frame.
755 * @param frame a frame to which the side data should be added
756 * @param type type of the added side data
757 * @param size size of the side data
759 * @return newly added side data on success, NULL on error
761 AVFrameSideData *av_frame_new_side_data(AVFrame *frame,
762 enum AVFrameSideDataType type,
766 * @return a pointer to the side data of a given type on success, NULL if there
767 * is no side data with such type in this frame.
769 AVFrameSideData *av_frame_get_side_data(const AVFrame *frame,
770 enum AVFrameSideDataType type);
773 * If side data of the supplied type exists in the frame, free it and remove it
776 void av_frame_remove_side_data(AVFrame *frame, enum AVFrameSideDataType type);
780 * Flags for frame cropping.
784 * Apply the maximum possible cropping, even if it requires setting the
785 * AVFrame.data[] entries to unaligned pointers. Passing unaligned data
786 * to Libav API is generally not allowed, and causes undefined behavior
787 * (such as crashes). You can pass unaligned data only to Libav APIs that
788 * are explicitly documented to accept it. Use this flag only if you
789 * absolutely know what you are doing.
791 AV_FRAME_CROP_UNALIGNED = 1 << 0,
795 * Crop the given video AVFrame according to its crop_left/crop_top/crop_right/
796 * crop_bottom fields. If cropping is successful, the function will adjust the
797 * data pointers and the width/height fields, and set the crop fields to 0.
799 * In all cases, the cropping boundaries will be rounded to the inherent
800 * alignment of the pixel format. In some cases, such as for opaque hwaccel
801 * formats, the left/top cropping is ignored. The crop fields are set to 0 even
802 * if the cropping was rounded or ignored.
804 * @param frame the frame which should be cropped
805 * @param flags Some combination of AV_FRAME_CROP_* flags, or 0.
807 * @return >= 0 on success, a negative AVERROR on error. If the cropping fields
808 * were invalid, AVERROR(ERANGE) is returned, and nothing is changed.
810 int av_frame_apply_cropping(AVFrame *frame, int flags);
813 * @return a string identifying the side data type
815 const char *av_frame_side_data_name(enum AVFrameSideDataType type);
821 #endif /* AVUTIL_FRAME_H */