【視頻編解碼·學習筆記】11. 提取SPS信息程序
一、準備工作:
回到之前SimpleH264Analyzer
程序,找到SPS信息,並對其做解析
調整項目目錄結構:
修改Global.h
文件中代碼,添加新數據類型UINT16,之前編寫的工程中,UINT8和UINT32都為小寫表示,為了更符合編程規範,將其改為全大寫(可使用ctrl+H在整個解決方案內進行替換)。
typedef unsigned char UINT8;
typedef unsigned short UINT16;
typedef unsigned int UINT32;
之後編寫的程序會有越來越多的輸出,如果全部輸入到控制臺中,會非常雜亂。因此輸出變成兩種方式,一種在控制臺輸出,另一種輸出到日誌文件中。步驟如下:
1 新建Configuration.h
文件,放到1.Application
目錄下,添加代碼:
#ifdef _CONFIGURATION_H_
#define _CONFIGURATION_H_
#include <fstream>
#define TRACE_CONFIG_CONSOLE 1
#define TRACE_CONFIG_LOGOUT 1
extern std::ofstream g_traceFile;
#endif
2 新建Configuration.cpp
,放到1.Application
目錄下,添加代碼:
#include "stdafx.h"
#include "Configuration.h"
#if TRACE_CONFIG_LOGOUT
std::ofstream g_traceFile;
#endif
3 在stdafx.h
中添加引用庫:
#include <string>
#include "Configuration.h"
4 是否寫入日誌文件定義在Stream.cpp
中的構造函數中:
在 CStreamFile::CStreamFile(TCHAR * fileName) 中添加:
#if TRACE_CONFIG_LOGOUT
g_traceFile.open(L"trace.txt");
if (!g_traceFile.is_open())
{
file_error(1 );
}
g_traceFile << "Trace file:" << endl;
#endif
析構函數CStreamFile::~CStreamFile()
中添加:
#ifdef TRACE_CONFIG_LOGOUT
if (g_traceFile.is_open())
{
g_traceFile.close();
}
#endif
當日誌文件打開失敗時,調用函數file_error(1),因此修改void CStreamFile::file_error(int idx) 函數,在其中添加錯誤代碼1的方案:
case 1:
wcout << L"Error: opening trace file failed." << endl;
break;
完成以上配置後編譯運行程序,在 \bin\Debug 目錄下會生成一個trace.txt
文件,寫入了這個字符串“Trace file:”
為了替換之前在控制臺直接輸出,在CStreamFile類中新建一個函數,首先在Stream.h文件中聲明函數(private)
void dump_NAL_type(UINT8 nalType);
在Stream.cpp
中添加這個函數的實現
void CStreamFile::dump_NAL_type(UINT8 nalType)
{
#if TRACE_CONFIG_CONSOLE
wcout << L"NAL Unit Type: " << nalType << endl;
#endif
#if TRACE_CONFIG_LOGOUT
g_traceFile << "NAL Unit Type: " << to_string(nalType) << endl;
#endif
}
將 Parse_h264_bitstream() 函數中 wcout輸出改為調用新函數:
dump_NAL_type(nalType);
重新編譯運行,由於此時控制臺和日誌文件輸出開關均打開,因此可在控制臺和trace.txt中看到NAL Unit Type的輸出
二、定義SPS類:
新建類CSeqParamSet,將生成的CSeqParamSet.h
、CSeqParamSet.cpp
放到 “3.NAL Unit” 目錄下
按照上一個筆記中官方文檔中提到的編碼結構,將所有語法元素一一定義出來,並設置setter函數:
修改SeqParamSet.h
:
#ifndef _SEQ_PARAM_SET_H_
#define _SEQ_PARAM_SET_H_
class CSeqParamSet
{
public:
CSeqParamSet();
~CSeqParamSet();
void Set_profile_level_idc(UINT8 profile, UINT8 level);
void Set_sps_id(UINT8 spsID);
void Set_chroma_format_idc(UINT8 chromaFormatIdc);
void Set_bit_depth(UINT8 bit_depth_luma, UINT8 bit_depth_chroma);
void Set_max_frame_num(UINT32 maxFrameNum);
void Set_poc_type(UINT8 pocType);
void Set_max_poc_cnt(UINT32 maxPocCnt);
void Set_max_num_ref_frames(UINT32 maxRefFrames);
void Set_sps_multiple_flags(UINT32 flags);
void Set_pic_reslution_in_mbs(UINT16 widthInMBs, UINT16 heightInMapUnits);
void Set_frame_crop_offset(UINT32 offsets[4]);
private:
UINT8 m_profile_idc;
UINT8 m_level_idc;
UINT8 m_sps_id;
// for uncommon profile...
UINT8 m_chroma_format_idc;
bool m_separate_colour_plane_flag;
UINT8 m_bit_depth_luma;
UINT8 m_bit_depth_chroma;
bool m_qpprime_y_zero_transform_bypass_flag;
bool m_seq_scaling_matrix_present_flag;
// ...for uncommon profile
UINT32 m_max_frame_num;
UINT8 m_poc_type;
UINT32 m_max_poc_cnt;
UINT32 m_max_num_ref_frames;
bool m_gaps_in_frame_num_value_allowed_flag;
UINT16 m_pic_width_in_mbs;
UINT16 m_pic_height_in_map_units;
UINT16 m_pic_height_in_mbs; // 圖像實際高度 not defined in spec, derived...
bool m_frame_mbs_only_flag;
bool m_mb_adaptive_frame_field_flag;
bool m_direct_8x8_inference_flag;
bool m_frame_cropping_flag;
UINT32 m_frame_crop_offset[4];
bool m_vui_parameters_present_flag;
// UINT32 m_reserved;
};
#endif
在SeqParamSet.cpp
文件中實現所有的setter函數,就是一個簡單的賦值過程:
#include "stdafx.h"
#include "SeqParamSet.h"
CSeqParamSet::CSeqParamSet()
{
}
CSeqParamSet::~CSeqParamSet()
{
}
void CSeqParamSet::Set_profile_level_idc(UINT8 profile, UINT8 level)
{
m_profile_idc = profile;
m_level_idc = level;
}
void CSeqParamSet::Set_sps_id(UINT8 sps_id)
{
m_sps_id = sps_id;
}
void CSeqParamSet::Set_chroma_format_idc(UINT8 chromaFormatIdc)
{
m_chroma_format_idc = chromaFormatIdc;
}
void CSeqParamSet::Set_bit_depth(UINT8 bit_depth_luma, UINT8 bit_depth_chroma)
{
m_bit_depth_luma = bit_depth_luma;
m_bit_depth_chroma = bit_depth_chroma;
}
void CSeqParamSet::Set_max_frame_num(UINT32 maxFrameNum)
{
m_max_frame_num = maxFrameNum;
}
void CSeqParamSet::Set_poc_type(UINT8 pocType)
{
m_poc_type = pocType;
}
void CSeqParamSet::Set_max_poc_cnt(UINT32 maxPocCnt)
{
m_max_poc_cnt = maxPocCnt;
}
void CSeqParamSet::Set_max_num_ref_frames(UINT32 maxRefFrames)
{
m_max_num_ref_frames = maxRefFrames;
}
void CSeqParamSet::Set_sps_multiple_flags(UINT32 flags)
{
m_separate_colour_plane_flag = flags & (1 << 21);
m_qpprime_y_zero_transform_bypass_flag = flags & (1 << 20);
m_seq_scaling_matrix_present_flag = flags & (1 << 19);
m_gaps_in_frame_num_value_allowed_flag = flags & (1 << 5);
m_frame_mbs_only_flag = flags & (1 << 4);
m_mb_adaptive_frame_field_flag = flags & (1 << 3);
m_direct_8x8_inference_flag = flags & (1 << 2);
m_frame_cropping_flag = flags & (1 << 1);
m_vui_parameters_present_flag = flags & 1;
}
void CSeqParamSet::Set_pic_reslution_in_mbs(UINT16 widthInMBs, UINT16 heightInMapUnits)
{
m_pic_width_in_mbs = widthInMBs;
m_pic_height_in_map_units = heightInMapUnits;
m_pic_height_in_mbs = m_frame_mbs_only_flag ? m_pic_height_in_map_units : 2 * m_pic_height_in_map_units;
}
void CSeqParamSet::Set_frame_crop_offset(UINT32 offsets[4])
{
for (int idx = 0; idx < 4; idx++)
{
m_frame_crop_offset[idx] = offsets[idx];
}
}
三、無符號指數哥倫布數據解碼:
與學習筆記9中實現的無符號指數哥倫布解碼部分完全相同,僅將代碼放在下面(筆記9中有詳細解釋):
在0.Global目錄下,新建Utils.h
,定義指數哥倫布編碼中兩個必要的函數:
#ifndef _UTILS_H_
#define _UTILS_H_
#include "Global.h"
int Get_bit_at_position(UINT8 *buf, UINT8 &bytePosition, UINT8 &bitPosition);
int Get_uev_code_num(UINT8 *buf, UINT8 &bytePosition, UINT8 &bitPosition);
#endif
在0.Global目錄下,新建Utils.cpp
,實現上面兩個函數:
#include "stdafx.h"
#include "Utils.h"
// 根據bytePosition和bitPosition 獲取當前比特位二進制數值 返回0/1
int Get_bit_at_position(UINT8 * buf, UINT8 & bytePosition, UINT8 & bitPosition)
{
UINT8 mask = 0, val = 0;
mask = 1 << (7 - bitPosition);
val = ((buf[bytePosition] & mask) != 0);
if (++bitPosition > 7)
{
bytePosition++;
bitPosition = 0;
}
return val;
}
// 將接下來一個指數哥倫布編碼 轉換成十進制數值
int Get_uev_code_num(UINT8 * buf, UINT8 & bytePosition, UINT8 & bitPosition)
{
assert(bitPosition < 8);
UINT8 val = 0, prefixZeroCount = 0;
int prefix = 0, surfix = 0;
while (true)
{
val = Get_bit_at_position(buf, bytePosition, bitPosition);
if (val == 0)
{
prefixZeroCount++;
}
else
{
break;
}
}
prefix = (1 << prefixZeroCount) - 1;
for (size_t i = 0; i < prefixZeroCount; i++)
{
val = Get_bit_at_position(buf, bytePosition, bitPosition);
surfix += val * (1 << (prefixZeroCount - i - 1));
}
prefix += surfix;
return prefix;
}
可將學習筆記9主函數中的代碼復制過來進行測試,能正確輸出解碼結果即可。
四、解析NALUnit中SPS數據:
將UALUnit中的語法元素,按照協議規定解析為SPS中各個成員變量的值
在NALUnit.h
和NALUnit.cpp
中添加函數,Parse_as_seq_param_set() 用於解析語法元素,代碼如下。(均按照學習筆記10中官方文檔順序解析即可)
int CNalUnit::Parse_as_seq_param_set(CSeqParamSet * sps)
{
UINT8 profile_idc = 0;
UINT8 level_idc = 0;
UINT8 sps_id = 0;
UINT8 chroma_format_idc = 0;
bool separate_colour_plane_flag = 0;
UINT8 bit_depth_luma = 0;
UINT8 bit_depth_chroma = 0;
bool qpprime_y_zero_transform_bypass_flag = 0;
bool seq_scaling_matrix_present_flag = 0;
UINT32 max_frame_num = 0;
UINT8 poc_type = 0;
UINT32 max_poc_cnt = 0;
UINT32 max_num_ref_frames = 0;
bool gaps_in_frame_num_value_allowed_flag = 0;
UINT16 pic_width_in_mbs = 0;
UINT16 pic_height_in_map_units = 0;
UINT16 pic_height_in_mbs = 0; // 圖像實際高度 not defined in spec, derived...
bool frame_mbs_only_flag = 0;
bool mb_adaptive_frame_field_flag = 0;
bool direct_8x8_inference_flag = 0;
bool frame_cropping_flag = 0;
UINT32 frame_crop_offset[4] = { 0 };
bool vui_parameters_present_flag = 0;
UINT8 bytePosition = 3, bitPosition = 0;
UINT32 flags = 0; //會檢索到各種flag元素,每個元素占一個比特,最終按先後順序放到flags中
profile_idc = m_pSODB[0];
// 第二個字節是constraint_set_flag 暫時用不到,空過去m_pSODB[1]
level_idc = m_pSODB[2];
sps_id = Get_uev_code_num(m_pSODB, bytePosition, bitPosition); //這裏是一個無符號指數哥倫布編碼,用前面寫好的函數提取
if (profile_idc == 100 || profile_idc == 110 || profile_idc == 122 || profile_idc == 244 || profile_idc == 44 ||
profile_idc == 83 || profile_idc == 86 || profile_idc == 118 || profile_idc == 128)
{
chroma_format_idc = Get_uev_code_num(m_pSODB, bytePosition, bitPosition);
if (chroma_format_idc == 3)
{
separate_colour_plane_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
// 提取到的單個flag,放到flag集合中的(可用的最高位上)
flags |= (separate_colour_plane_flag << 21);
}
bit_depth_luma = Get_uev_code_num(m_pSODB, bytePosition, bitPosition) + 8;
bit_depth_chroma = Get_uev_code_num(m_pSODB, bytePosition, bitPosition) + 8;
qpprime_y_zero_transform_bypass_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= (qpprime_y_zero_transform_bypass_flag << 20);
seq_scaling_matrix_present_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= (seq_scaling_matrix_present_flag << 19);
if (seq_scaling_matrix_present_flag)
{
// 這個部分暫時用不到,先返回一個錯誤碼代替
return -1;
}
}
// 下面不求log2_max_frame_num,而是直接將原來的數字求出來
max_frame_num = 1 << (Get_uev_code_num(m_pSODB, bytePosition, bitPosition) + 4);
poc_type = Get_uev_code_num(m_pSODB, bytePosition, bitPosition);
if (0 == poc_type)
{
max_poc_cnt = 1 << (Get_uev_code_num(m_pSODB, bytePosition, bitPosition) + 4);
}
else
{
// 暫時不考慮這種情況
return -1;
}
max_num_ref_frames = Get_uev_code_num(m_pSODB, bytePosition, bitPosition);
gaps_in_frame_num_value_allowed_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= (gaps_in_frame_num_value_allowed_flag << 5); //中間跳過了好多位,為本該有卻沒實現的flag留出位置
pic_width_in_mbs = Get_uev_code_num(m_pSODB, bytePosition, bitPosition) + 1;
pic_height_in_map_units = Get_uev_code_num(m_pSODB, bytePosition, bitPosition) + 1;
frame_mbs_only_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= (frame_mbs_only_flag << 4);
if (!frame_mbs_only_flag)
{
mb_adaptive_frame_field_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= (mb_adaptive_frame_field_flag << 3);
}
direct_8x8_inference_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= (direct_8x8_inference_flag << 2);
frame_cropping_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= (direct_8x8_inference_flag << 1);
if (frame_cropping_flag)
{
for (int idx = 0; idx < 4; idx++)
{
frame_crop_offset[idx] = Get_uev_code_num(m_pSODB, bytePosition, bitPosition);
}
}
vui_parameters_present_flag = Get_bit_at_position(m_pSODB, bytePosition, bitPosition);
flags |= vui_parameters_present_flag;
// 解析碼流完成
sps->Set_profile_level_idc(profile_idc, level_idc);
sps->Set_sps_id(sps_id);
sps->Set_chroma_format_idc(chroma_format_idc);
sps->Set_bit_depth(bit_depth_luma, bit_depth_chroma);
sps->Set_max_frame_num(max_frame_num);
sps->Set_poc_type(poc_type);
sps->Set_max_poc_cnt(max_poc_cnt);
sps->Set_max_num_ref_frames(max_num_ref_frames);
sps->Set_sps_multiple_flags(flags);
sps->Set_pic_reslution_in_mbs(pic_width_in_mbs, pic_height_in_map_units);
if (frame_cropping_flag)
{
sps->Set_frame_crop_offset(frame_crop_offset);
}
return 0;
}
五、添加調用部分:
回到Stream.cpp
中,找到Parse_h264_bitstream() 函數,在學習筆記6中,已經完成了nalType的提取,並得到了SODB數據,在後面添加解析序列參數集sps的部分。
CNalUnit nalUint(&m_nalVec[1], m_nalVec.size() - 1);
switch (nalType)
{
case 7:
// 解析SPS NAL 數據
if (m_sps)
{
delete m_sps;
}
m_sps = new CSeqParamSet;
nalUint.Parse_as_seq_param_set(m_sps);
break;
default:
break;
}
可對其進行單步調試,重點看這兩個參數 pic_width_in_mbs,pic_height_in_map_units,分別是以宏塊為單位的寬、高分辨率。本次調試使用的視頻仍是學習筆記3使用的視頻,之前設置的參數為:
SourceWidth = 176 # Image width in Pels, must be multiple of 16
SourceHeight = 144 # Image height in Pels, must be multiple of 16
宏塊分辨率要在原來基礎上除16,即寬11、高9。這兩個參數吻合,基本表明程序沒有問題。
【視頻編解碼·學習筆記】11. 提取SPS信息程序