1 /* ------------------------------------------------------------------ 2 * Copyright (C) 1998-2009 PacketVideo 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either 13 * express or implied. 14 * See the License for the specific language governing permissions 15 * and limitations under the License. 16 * ------------------------------------------------------------------- 17 */ 18 #ifndef _MP4ENC_API_H_ 19 #define _MP4ENC_API_H_ 20 21 #include <string.h> 22 23 #ifndef _PV_TYPES_ 24 #define _PV_TYPES_ 25 typedef unsigned char UChar; 26 typedef char Char; 27 typedef unsigned int UInt; 28 typedef int Int; 29 typedef unsigned short UShort; 30 typedef short Short; 31 typedef unsigned int Bool; 32 typedef uint32_t ULong; 33 34 #define PV_CODEC_INIT 0 35 #define PV_CODEC_STOP 1 36 #endif 37 38 #define PV_TRUE 1 39 #define PV_FALSE 0 40 41 typedef enum 42 { 43 SHORT_HEADER, 44 SHORT_HEADER_WITH_ERR_RES, 45 H263_MODE, 46 H263_MODE_WITH_ERR_RES, 47 DATA_PARTITIONING_MODE, 48 COMBINE_MODE_NO_ERR_RES, 49 COMBINE_MODE_WITH_ERR_RES 50 51 } MP4EncodingMode; 52 53 typedef enum 54 { 55 CONSTANT_Q, 56 CBR_1, 57 VBR_1, 58 CBR_2, 59 VBR_2, 60 CBR_LOWDELAY 61 } MP4RateControlType; 62 63 typedef enum 64 { 65 PASS1, 66 PASS2 67 } PassNum; 68 69 typedef enum 70 { 71 PV_OFF, 72 PV_ON 73 } ParamEncMode; 74 75 76 /* {SPL0, SPL1, SPL2, SPL3, CPL1, CPL2, CPL2, CPL2} , SPL0: Simple Profile@Level0 , CPL1: Core Profile@Level1 */ 77 /* {SSPL0, SSPL1, SSPL2, SSPL2, CSPL1, CSPL2, CSPL3, CSPL3} , SSPL0: Simple Scalable Profile@Level0, CPL1: Core Scalable Profile@Level1 */ 78 79 typedef enum 80 { 81 /* Non-scalable profile */ 82 SIMPLE_PROFILE_LEVEL0 = 0, 83 SIMPLE_PROFILE_LEVEL1, 84 SIMPLE_PROFILE_LEVEL2, 85 SIMPLE_PROFILE_LEVEL3, 86 CORE_PROFILE_LEVEL1, 87 CORE_PROFILE_LEVEL2, 88 89 /* Scalable profile */ 90 SIMPLE_SCALABLE_PROFILE_LEVEL0 = 6, 91 SIMPLE_SCALABLE_PROFILE_LEVEL1, 92 SIMPLE_SCALABLE_PROFILE_LEVEL2, 93 94 CORE_SCALABLE_PROFILE_LEVEL1 = 10, 95 CORE_SCALABLE_PROFILE_LEVEL2, 96 CORE_SCALABLE_PROFILE_LEVEL3 97 98 } ProfileLevelType; 99 100 101 typedef struct tagMP4HintTrack 102 { 103 UChar MTB; 104 UChar LayerID; 105 UChar CodeType; 106 UChar RefSelCode; 107 } MP4HintTrack; 108 109 typedef struct tagvideoEncControls 110 { 111 void *videoEncoderData; 112 Int videoEncoderInit; 113 } VideoEncControls; 114 115 116 typedef struct tagvideoEncFrameIO 117 { 118 UChar *yChan; /* pointer to Y */ 119 UChar *uChan; /* pointer to U */ 120 UChar *vChan; /* pointer to V */ 121 Int height; /* height for Y */ 122 Int pitch; /* stride for Y */ 123 ULong timestamp; /* modulo timestamp in millisecond*/ 124 125 } VideoEncFrameIO ; 126 127 /** 128 @brief Encoding options structure */ 129 typedef struct tagvideoEncOptions 130 { 131 /** @brief Sets the encoding mode, defined by the above enumaration. If there are conflicts between the encoding mode 132 * and subsequent encoding options, encoding mode take precedent over encoding options. */ 133 MP4EncodingMode encMode; 134 135 /** @brief Sets the number of bytes per packet, only used in DATA_PARTITIONING_MODE or COMBINE_MODE_WITH_ERR_RES mode. 136 * The resync marker will be inserted as often as the size of the packet.*/ 137 Int packetSize; 138 139 /** @brief Selects MPEG-4/H.263 profile and level, if specified other encoding options must conform with it. */ 140 ProfileLevelType profile_level; 141 142 /** @brief Enables reversible variable length code (RVLC) mode. Normally it is set to PV_OFF.*/ 143 ParamEncMode rvlcEnable; 144 145 /** @brief Set the frequency of GOB header interval */ 146 Int gobHeaderInterval; 147 148 /** @brief Sets the number of bitstream layers: 1 is base only: 2 is base + enhancement */ 149 Int numLayers; 150 151 /** @brief Sets the number of ticks per second used for timing information encoded in MPEG4 bitstream.*/ 152 Int timeIncRes; 153 154 /** @brief Sets the number of ticks in time increment resolution between 2 source frames (equivalent to source frame rate). */ 155 Int tickPerSrc; 156 157 /** @brief Specifies encoded heights in pixels, height[n] represents the n-th layer's height. */ 158 Int encHeight[2]; 159 160 /** @brief Specifies encoded widths in pixels, width[n] represents the n-th layer's width.*/ 161 Int encWidth[2]; 162 163 /** @brief Specifies target frame rates in frames per second, frameRate[n] represents the n-th layer's target frame rate.*/ 164 float encFrameRate[2]; 165 166 /** @brief Specifies target bit rates in bits per second unit, bitRate[n] represents the n-th layer's target bit rate. */ 167 Int bitRate[2]; 168 169 /** @brief Specifies default quantization parameters for I-Vop. Iquant[n] represents the n-th layer default quantization parameter. The default is Iquant[0]=12.*/ 170 Int iQuant[2]; 171 172 /** @brief Specifies default quantization parameters for P-Vop. Pquant[n] represents the n-th layer default quantization parameter. The default is Pquant[0]=10.*/ 173 Int pQuant[2]; 174 175 /** @brief specifies quantization mode (H263 mode or MPEG mode) of the encoded base and enhance layer (if any). 176 * In Simple and Simple Scalable profile, we use only H263 mode.*/ 177 Int quantType[2]; 178 179 /** @brief Sets rate control algorithm, one of (CONSTANT_Q, CBR_1, or VBR_1). 180 * CONSTANT_Q uses the default quantization values to encode the sequence. 181 * CBR_1 (constant bit rate) controls the output at a desired bit rate 182 * VBR_1 (variable bit rate) gives better picture quality at the expense of bit rate fluctuation 183 * Note: type=CONSTANT_Q produces sequences with arbitrary bit rate. 184 * type=CBR_1 produces sequences suitable for streaming. 185 * type=VBR_1 produces sequences suitable for download. */ 186 MP4RateControlType rcType; 187 188 /** @brief Sets the VBV buffer size (in the unit of second delay) used to prevent buffer overflow and underflow 189 * on the decoder side. This function is redundant to PVSetVBVSize. Either one of them is used at a time. */ 190 float vbvDelay; 191 192 /** @brief Specifies whether frame skipping is permitted or not. When rate control type is set to CONSTANT_Q 193 * frame skipping is automatically banned. In CBR_1 and VBR_1 rate control, frame skipping is allowed by default. 194 * However, users can force no frame skipping with this flag, but buffer constraint may be violated.*/ 195 ParamEncMode noFrameSkipped; 196 197 /** @brief Sets the maximum number of P-frames between two I-frames. I-frame mode is periodically forced 198 * if no I-frame is encoded after the specified period to add error resiliency and help resynchronize in case of errors. 199 * If scene change detection can add additional I-frame if new scenes are detected. 200 * intraPeriod is the I frame interval in terms of second. 201 * intraPeriod =0 indicates I-frame encoding only; 202 * intraPeriod = -1 indicates I-frame followed by all P-frames; (default) 203 * intraPeriod = N, indicates the number of P-frames between 2 I-frames.*/ 204 Int intraPeriod; 205 206 207 /** @brief Specifies the number Intra MBs to be refreshed in a P-frame. */ 208 Int numIntraMB; 209 210 /** 211 * @brief Specifies whether the scene change detection (SCD) is enabled or disabled. 212 * With SCD enable, when a new scene is detected, I-Vop mode will be used for the first frame of 213 * the new scene resulting in better picture quality. An insertion of an I-VOP resets the intraPeriod 214 * specified by the IntraPeriodAPI().*/ 215 ParamEncMode sceneDetect; 216 217 /** @brief Specifies the search range of motion estimation search. Larger value implies 218 * larger search range, better motion vector match, but more complexity. 219 * If searchRange=n, the motion vector search is in the range of [-n,n-1] pixels. 220 * If half-pel mode is on, the range is [-n, (n-1)+1/2] pixels. The default value is 16.*/ 221 Int searchRange; 222 223 /** @brief Turns on/off 8x8 block motion estimation and compensation. 224 * If on, four motion vectors may be used for motion estimation and compensation of a macroblock, 225 * otherwise one motion vector per macroblock is used. When the 8x8 MV is off, the total encoding complexity 226 * is less but the image quality is also worse. Therefore, it can be used in complexity limited environment.*/ 227 ParamEncMode mv8x8Enable; 228 229 230 /** @brief Set the threshold for using intra DC VLC. 231 * Value must range from 0-7.*/ 232 Int intraDCVlcTh; 233 234 /** @brief This flag turns on the use of AC prediction */ 235 Bool useACPred; 236 237 } VideoEncOptions; 238 239 #ifdef __cplusplus 240 extern "C" 241 { 242 #endif 243 244 245 /* API's */ 246 /* Always start with this one !!*/ 247 /** 248 * @brief Gets default encoding options. This way users only have to set relevant encoding options and leave the one 249 * they are unsure of. 250 * @encOption Pointer to VideoEncOption structure. 251 * @encUseCase This value determines the set of default encoding options, for example, different encoding options 252 * are assigned to streaming use-case as compared to download use-case. It can be project dependent too. 253 * @return true for correct operation; false if error happens 254 */ 255 OSCL_IMPORT_REF Bool PVGetDefaultEncOption(VideoEncOptions *encOption, Int encUseCase); 256 257 /** 258 * @brief Verifies the consistency of encoding parameters, allocates memory needed and set necessary internal variables. 259 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 260 * @return true for correct operation; false if error happens 261 */ 262 OSCL_IMPORT_REF Bool PVInitVideoEncoder(VideoEncControls *encCtrl, VideoEncOptions *encOption); 263 264 /* acquiring encoder info APIs */ 265 /** 266 * @brief This function returns VOL header. It has to be called before the frame is encoded. If so, 267 * then the VOL Header is passed back to the application. Then all frames that are encoded do not contain the VOL Header. 268 * If you do not call the API then the VOL Header is passed within the first frame that is encoded. 269 * The behavior is unknown if it is called after the first frame is encoded. It is mainly used for MP4 file format authoring. 270 * @param encCtrl is video encoder control structure that is always passed as input in all APIs. 271 * @param volHeader is the Buffer for VOL header. 272 * @param size is the size of VOL header in bytes. 273 * @param layer is the layer of the requested VOL header. 274 * @return true for correct operation; false if error happens. 275 */ 276 OSCL_IMPORT_REF Bool PVGetVolHeader(VideoEncControls *encCtrl, UChar *volHeader, Int *size, Int layer); 277 278 /** 279 * @brief This function returns the profile and level in H.263 coding when the encoding parameters are set 280 * @param encCtrl is video encoder control structure that is always passed as input in all APIs. 281 * @param profileID is the pointer of the profile ID. Right now we only support profile 0 282 * @param levelID is the pointer of the level ID that could be 10-70. 283 * @return true for correct operation; false if error happens. 284 */ 285 OSCL_IMPORT_REF Bool PVGetH263ProfileLevelID(VideoEncControls *encCtrl, Int *profileID, Int *levelID); 286 287 /** 288 * @brief This function returns the profile and level of MPEG4 when the encoding parameters are set 289 * @param encCtrl is video encoder control structure that is always passed as input in all APIs. 290 * @param profile_level is the pointer of the profile enumeration 291 * @param nLayer is the index of the layer of interest 292 * @return true for correct operation; false if error happens. 293 */ 294 OSCL_IMPORT_REF Bool PVGetMPEG4ProfileLevelID(VideoEncControls *encCtrl, Int *profile_level, Int nLayer); 295 296 /** 297 * @brief This function returns maximum frame size in bytes 298 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 299 * @param maxVideoFrameSize is the pointer of the maximum frame size 300 * @return true for correct operation; false if error happens 301 */ 302 OSCL_IMPORT_REF Bool PVGetMaxVideoFrameSize(VideoEncControls *encCtrl, Int *maxVideoFrameSize); 303 304 #ifndef LIMITED_API 305 /** 306 * @brief This function returns the total amount of memory (in bytes) allocated by the encoder library. 307 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 308 * @return true for correct operation; false if error happens 309 */ 310 OSCL_IMPORT_REF Int PVGetEncMemoryUsage(VideoEncControls *encCtrl); 311 312 /** 313 * @brief This function is used by PVAuthor to get the size of the VBV buffer. 314 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 315 * @param VBVSize is the pointer of The size of the VBV buffer in bytes. 316 * @return true for correct operation; false if error happens 317 */ 318 OSCL_IMPORT_REF Bool PVGetVBVSize(VideoEncControls *encCtrl, Int *VBVSize); 319 #endif 320 321 /** 322 * @brief This function encodes a frame in YUV 4:2:0 format from the *video_in input frame and put the result in YUV 323 * for reconstructed frame and bstream for MPEG4 bitstream. The application is required to allocate memory for 324 * bitstream buffer.The size of the input bitstream memory and the returned output buffer are specified in the 325 * size field. The encoded layer is specified by the nLayer field. If the current frame is not encoded, size=0 and nLayer=-1. 326 * Note: If the allocated buffer size is too small to fit a bitstream of a frame, then those extra bits will be left out 327 * which can cause syntactic error at the decoder side. 328 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 329 * @param vid_in is the pointer to VideoEncFrameIO structure containing the YUV input data 330 * @param vid_out is the pointer to VideoEncFrameIO structure containing the reconstructed YUV output data after encoding 331 * @param nextModTime is the timestamp encoder expects from the next input 332 * @param bstream is the pointer to MPEG4 bitstream buffer 333 * @param size is the size of bitstream buffer allocated (input) and size of the encoded bitstream (output). 334 * @param nLayer is the layer of the encoded frame either 0 for base or 1 for enhancement layer. The value -1 indicates skipped frame due to buffer overflow. 335 * @return true newfor correct operation; false if error happens 336 */ 337 OSCL_IMPORT_REF Bool PVEncodeVideoFrame(VideoEncControls *encCtrl, VideoEncFrameIO *vid_in, VideoEncFrameIO *vid_out, 338 ULong *nextModTime, UChar *bstream, Int *size, Int *nLayer); 339 340 341 /** 342 * @brief This function is used to query overrun buffer. It is used when PVEncodeVideoFrame.returns size that is 343 * larger than the input size. 344 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 345 * @return Pointer to the overrun buffer. NULL if overrun buffer is not used. 346 */ 347 OSCL_IMPORT_REF UChar* PVGetOverrunBuffer(VideoEncControls *encCtrl); 348 349 #ifndef NO_SLICE_ENCODE /* This set of APIs are not working. This functionality has been partially 350 replaced by the introduction of overrun buffer. */ 351 352 /* slice-based coding */ 353 /** 354 * @brief This function sets the input YUV frame and timestamp to be encoded by the slice-based encoding function PVEncodeSlice(). 355 * It also return the memory address the reconstructed frame will be copied to (in advance) and the coded layer number. 356 * The encoder library processes the timestamp and determine if this frame is to be encoded or not. If the current frame 357 * is not encoded, nLayer=-1. For frame-based motion estimation, the motion estimation of the entire frame is also performed 358 * in this function. For MB-based motion estimation, the motion vector is searched while coding each MB in PVEncodeSlice(). 359 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 360 * @param vid_in is the pointer to VideoEncFrameIO structure containing the YUV input data 361 * @param nextModTime is the timestamp encoder expects from the next input if this input is rejected and nLayer is set to -1. 362 * @param nLayer is the layer of the encoded frame either 0 for base or 1 for enhancement layer. The value -1 indicates skipped frame due to buffer overflow. 363 * @return true newfor correct operation; false if error happens 364 */ 365 OSCL_IMPORT_REF Bool PVEncodeFrameSet(VideoEncControls *encCtrl, VideoEncFrameIO *vid_in, ULong *nextModTime, Int *nLayer); 366 /** 367 * @brief This function encodes a GOB (short header mode) or a packet (data partitioning mode or combined mode with resync marker) 368 * and output the reconstructed frame and MPEG4 bitstream. The application is required to allocate memory for the bitstream buffer. 369 * The size of the input bitstream memory and the returned output buffer are specified in the size field. If the buffer size is 370 * smaller than the requested packet size, user has to call PVEncodeSlice again to get the rest of that pending packet before moving 371 * on to the next packet. For the combined mode without resync marker, the function returns when the buffer is full. 372 * The end-of-frame flag indicates the completion of the frame encoding. Next frame must be sent in with PVEncodeFrameSet(). 373 * At the end-of-frame, the next video input address and the next video modulo timestamp will be set. 374 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 375 * @param bstream is the pointer to MPEG4 bitstream buffer. 376 * @param size is the size of bitstream buffer allocated (input) and size of the encoded bitstream (output). 377 * @param endofFrame is a flag indicating the end-of-frame, '1'. Otherwise, '0'. When PVSetNoCurrentFrameSkip is OFF, 378 * end-of-frame '-1' indicates current frame bitstream must be disregarded. 379 * @param vid_out is the pointer to VideoEncFrameIO structure containing the reconstructed YUV output data after encoding 380 * @param nextModTime is the timestamp encoder expects from the next input 381 * @return true newfor correct operation; false if error happens 382 */ 383 OSCL_IMPORT_REF Bool PVEncodeSlice(VideoEncControls *encCtrl, UChar *bstream, Int *size, 384 Int *endofFrame, VideoEncFrameIO *vid_out, ULong *nextModTime); 385 #endif 386 387 /** 388 * @brief This function returns MP4 file format hint track information. 389 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 390 * @param info is the structure for MP4 hint track information 391 * @return true for correct operation; false if error happens 392 */ 393 OSCL_IMPORT_REF Bool PVGetHintTrack(VideoEncControls *encCtrl, MP4HintTrack *info); 394 395 #ifndef LIMITED_API 396 /** 397 * @brief updates target frame rates of the encoded base and enhance layer (if any) while encoding operation is ongoing. 398 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 399 * @param frameRate is the pointers to array of target frame rates in frames per second, 400 * frameRate[n] represents the n-th layer's target frame rate. 401 * @return true for correct operation; false if error happens 402 */ 403 OSCL_IMPORT_REF Bool PVUpdateEncFrameRate(VideoEncControls *encCtrl, float *frameRate); /* for 2-way */ 404 405 406 /** 407 * @brief updates target bit rates of the encoded base and enhance layer (if any) while encoding operation is ongoing. 408 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 409 * @param bitRate is the pointers to array of target bit rates in bits per second unit, 410 * bitRate[n] represents the n-th layer's target bit rate. 411 * @return true for correct operation; false if error happens 412 */ 413 OSCL_IMPORT_REF Bool PVUpdateBitRate(VideoEncControls *encCtrl, Int *bitRate); /* for 2-way */ 414 415 416 /** 417 * @brief updates the INTRA frame refresh interval while encoding operation is ongoing. 418 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 419 * @param aIFramePeriod is a new value of INTRA frame interval in the unit of number of coded frames. 420 * @return true for correct operation; false if error happens 421 */ 422 423 OSCL_IMPORT_REF Bool PVUpdateIFrameInterval(VideoEncControls *encCtrl, Int aIFramePeriod);/* for 2-way */ 424 425 /** 426 * @brief specifies the number Intra MBs to be refreshed 427 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 428 * @param numMB is the number of Intra MBs to be refreshed 429 * @return true for correct operation; false if error happens 430 */ 431 OSCL_IMPORT_REF Bool PVUpdateNumIntraMBRefresh(VideoEncControls *encCtrl, Int numMB); /* for 2-way */ 432 433 /** 434 * @brief This function is called whenever users want the next base frame to be encoded as an I-Vop. 435 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 436 * @return true for correct operation; false if error happens 437 */ 438 OSCL_IMPORT_REF Bool PVIFrameRequest(VideoEncControls *encCtrl); /* for 2-way */ 439 440 #endif // LIMITED_API 441 442 /* finishing encoder */ 443 /** 444 * @brief This function frees up all the memory allocated by the encoder library. 445 * @param encCtrl is video encoder control structure that is always passed as input in all APIs 446 * @return true for correct operation; false if error happens 447 */ 448 OSCL_IMPORT_REF Bool PVCleanUpVideoEncoder(VideoEncControls *encCtrl); 449 450 #ifdef __cplusplus 451 } 452 #endif 453 #endif /* _MP4ENC_API_H_ */ 454 455
