快捷導(dǎo)航

C/C++實(shí)現(xiàn)H264文件解析

更新時(shí)間：2024年10月28日 10:13:43 作者：靜止了所有的花開

這篇文章主要為大家詳細(xì)介紹了如何通過(guò)C++實(shí)現(xiàn)H264文件以及一段H264碼流解析,文中的示例代碼講解詳細(xì),感興趣的小伙伴可以跟隨小編一起學(xué)習(xí)一下

H264視頻編碼格式簡(jiǎn)介

H264是視頻壓縮編碼標(biāo)準(zhǔn)。視頻?件的傳輸是?個(gè)極?的問(wèn)題：?段分辨率為19201080的視頻，每個(gè)像素點(diǎn)為RGB占?3個(gè)字節(jié)，幀率是25的視頻，對(duì)于傳輸帶寬的要求是：192010803*25/1024/1024=148.315MB/s，換成bps則意味著視頻每秒帶寬為1186.523Mbps，這樣的速率對(duì)于?絡(luò)存儲(chǔ)是不可接受的。因此視頻壓縮和編碼技術(shù)應(yīng)運(yùn)??。

H264碼流分析

通常我們使用H264碼流分析工具時(shí)，我們會(huì)發(fā)現(xiàn)H264碼流是由連續(xù)的幀組成的，一組連續(xù)的幀成為1個(gè)GOP即1組，對(duì)于1個(gè)GOP來(lái)說(shuō)，通常包括SPS（序列參數(shù)集，編碼視頻序列的全局參數(shù)）PPS（圖像參數(shù)集，?個(gè)序列中某?幅圖像或者某?幅圖像的參數(shù)），IDR幀(第一個(gè)出現(xiàn)的I幀成為IDR幀，1個(gè)GOP中通常只包含1個(gè)IDR幀)，P幀（GOP中剩余幀基本都是P幀）；GOP描述如下：

H264碼流結(jié)構(gòu)如圖：

C/C++ H264文件解析

C++實(shí)現(xiàn)H264文件以及一段H264碼流解析，源碼如下：

h264Parse.h:

#ifndef _H264PARSE_H_
#define _H264PARSE_H_
 
#include <fstream>
 
class H264Parse
{
public:
    int open_file(const std::string &filename);
 
    /**
     * @brief 從文件中讀取一個(gè)nalu，包含起始碼
     * @param buf 存放nalu的緩沖區(qū)
     * @param size 緩沖區(qū)大小
     * @param len nalu的長(zhǎng)度
     * @param n 每次讀取多少個(gè)字節(jié)
     * @return  -1 失敗    0 已到文件末尾     1 成功獲取到一個(gè)nalu
     */
    int read_nalu(uint8_t *buf, uint32_t size, uint32_t &len, uint32_t n);
 
    void close_file();
 
    // 獲取起始碼長(zhǎng)度
    static int get_startCode_len(const uint8_t *ptr);
 
    static const uint8_t *find_startCode_pos(const uint8_t *buf, uint32_t len);
 
    /**
     * @brief 從一段h264碼流中分割nalu，包含起始碼
     * @param stream h264碼流
     * @param streamLen 碼流大小
     * @param nalu Pointer to the extracted nalu
     * @param naluLen nalu的長(zhǎng)度
     * @param record Pointer用于記錄狀態(tài)，第一次分割時(shí)把 *record 賦值為NULL
     * @return  -1 失敗    0 已分割完     1 成功獲取到一個(gè)nalu
     */
    static int nalu_tok(const uint8_t *stream, uint32_t streamLen, const uint8_t **nalu,
                        uint32_t &naluLen, const uint8_t **record);
 
private:
    std::fstream h264File;
 
    int read_start_code(uint8_t *buf);
    int adjust_filePointer_pos(uint32_t totalRead, uint32_t naluLen);
};
 
#endif // _H264PARSE_H_

h264Parse.cpp:

#include "h264Parse.h"
#include <iostream>
#include <cstring>
 
int H264Parse::open_file(const std::string &filename)
{
    h264File.open(filename, std::ios::in | std::ios::binary);
    if (!h264File.is_open())
    {
        std::cout << "Failed to open the H.264 file." << std::endl;
        return -1;
    }
 
    return 0;
}
 
int H264Parse::get_startCode_len(const uint8_t *ptr)
{
    if (ptr[0] == 0x00 && ptr[1] == 0x00)
    {
        if (ptr[2] == 0x01)
            return 3;
        else if (ptr[2] == 0x00 && ptr[3] == 0x01)
            return 4;
    }
    return -1; // 無(wú)效的起始碼
}
 
// 讀取起始碼，并返回其長(zhǎng)度
int H264Parse::read_start_code(uint8_t *buf)
{
    // 讀取前4個(gè)字節(jié)來(lái)判斷起始碼長(zhǎng)度
    h264File.read(reinterpret_cast<char *>(buf), 4);
    if (h264File.gcount() < 4)
    {
        return -1;
    }
 
    return get_startCode_len(buf);
}
 
// 尋找NALU的起始碼位置
const uint8_t *H264Parse::find_startCode_pos(const uint8_t *buf, uint32_t len)
{
    const uint8_t *p = buf;
 
    if (len < 3)
        return NULL;
 
    for (uint32_t i = 0; i < len - 3; ++i)
    {
        if ((p[0] == 0x00 && p[1] == 0x00 && p[2] == 0x01) ||
            (p[0] == 0x00 && p[1] == 0x00 && p[2] == 0x00 && p[3] == 0x01))
        {
            return p;
        }
        p++;
    }
 
    // 檢查最后3字節(jié)是不是起始碼
    if (p[0] == 0x00 && p[1] == 0x00 && p[2] == 0x01)
        return p;
 
    return NULL;
}
 
// 調(diào)整文件指針位置
int H264Parse::adjust_filePointer_pos(uint32_t totalRead, uint32_t naluLen)
{
    int offset = -(totalRead - naluLen);
 
    if (!h264File.eof())
    {
        h264File.seekg(offset, std::ios::cur);
    }
    else
    {
        h264File.clear(); // 達(dá)到文件末尾了要先清除 eof 標(biāo)志
        h264File.seekg(offset, std::ios::end);
    }
 
    if (h264File.fail())
    {
        std::cout << "seekg failed!" << std::endl;
        return -1;
    }
 
    return 0;
}
 
int H264Parse::read_nalu(uint8_t *buf, uint32_t size, uint32_t &len, uint32_t n)
{
    uint32_t totalRead = 0;
 
    int startCodeLength = read_start_code(buf);
    if (startCodeLength == -1)
    {
        printf("read_start_code failed.\n");
        return -1;
    }
 
    totalRead += 4; // 已經(jīng)讀取了4字節(jié)的長(zhǎng)度
 
    while (true)
    {
        if (size < totalRead + n)
        {
            std::cout << "Buffer size is too small: size=" << size
                      << ", needed=" << totalRead + n << std::endl;
            return -1;
        }
 
        h264File.read(reinterpret_cast<char *>(buf + totalRead), n);
        std::streamsize bytesRead = h264File.gcount();
        if (bytesRead <= 0)
        {
            std::cout << "Failed to read from file!" << std::endl;
            return -1;
        }
 
        uint32_t searchStart = (totalRead > 4) ? totalRead - 3 : startCodeLength;
        const uint8_t *naluEnd = find_startCode_pos(buf + searchStart,
                                                    bytesRead + (totalRead > 4 ? 3 : 0));
        totalRead += bytesRead;
        if (naluEnd != nullptr)
        {
            len = naluEnd - buf;
            if (adjust_filePointer_pos(totalRead, len) < 0)
                return -1;
 
            break;
        }
 
        // 是否讀取到文件末尾
        if (h264File.peek() == std::char_traits<char>::eof())
        {
            len = totalRead;
            return 0; // NALU完整讀取
        }
    }
 
    memset(buf + len, 0, size - len); // 清空剩余部分
    return 1;                         // 成功讀取
}
 
void H264Parse::close_file()
{
    h264File.close();
}
 
int H264Parse::nalu_tok(const uint8_t *stream, uint32_t streamLen, const uint8_t **nalu,
                        uint32_t &naluLen, const uint8_t **record)
{
    const uint8_t *current = (record && *record) ? *record : stream;
    uint32_t offset = static_cast<uint32_t>(current - stream);
 
    if (offset >= streamLen)
    {
        return -1; // 當(dāng)前記錄位置超出緩沖區(qū)
    }
 
    int scLen = get_startCode_len(current);
    if (scLen == -1 || (current + scLen) > (stream + streamLen))
    {
        return -1; // 無(wú)效的起始碼或起始碼長(zhǎng)度超出緩沖區(qū)
    }
 
    // 查找下一個(gè)起始碼的位置
    const uint8_t *next_start = find_startCode_pos(current + scLen, streamLen - offset - scLen);
 
    if (next_start)
    {
        *nalu = current;
        naluLen = static_cast<uint32_t>(next_start - current);
        *record = next_start;
        return 1; // 成功獲取到一個(gè) NALU
    }
    else
    {
        // 最后一個(gè) NALU
        *nalu = current;
        naluLen = streamLen - offset;
        *record = NULL; // 重置記錄指針
        return 0;       // 分割完畢
    }
}

測(cè)試：

#include <iostream>
#include <vector>
#include "h264Parse.h"
 
void test1()
{
    int ret;
    int number = 0;
    H264Parse h264;
    uint8_t buf[1024 * 1024];
    uint32_t len = 0;
 
    h264.open_file("/home/tl/work/app/res/output.h264");
    while ((ret = h264.read_nalu(buf, sizeof(buf), len, 1024 * 2)) != -1)
    {
        printf("number: %d nalu len: %u\n", number, len - h264.get_startCode_len(buf));
        number++;
        if (ret == 0)
            break;
    }
    if (ret == -1)
    {
        std::cout << "read_nalu failed." << std::endl;
    }
    h264.close_file();
}
 
// 輔助函數(shù)：打印 NALU 信息
void print_nalu(const uint8_t *nalu, uint32_t len, int index)
{
    std::cout << "NALU " << index << ": Length = " << len << " bytes, Data = ";
    for (uint32_t i = 0; i < len; ++i)
    {
        printf("%02X ", nalu[i]);
    }
    std::cout << std::endl;
}
 
void test2()
{
    // 構(gòu)造一個(gè)模擬的 H.264 碼流緩沖區(qū)，包含多個(gè) NALU
    // 起始碼格式：0x000001 (3 字節(jié)) 和 0x00000001 (4 字節(jié))
    // NALU 內(nèi)容：隨機(jī)填充的字節(jié)數(shù)據(jù)
    std::vector<uint8_t> buffer;
 
    // NALU 1: 3 字節(jié)起始碼 + 5 字節(jié)數(shù)據(jù)
    std::vector<uint8_t> nalu1 = {0x00, 0x00, 0x01, 0x65, 0x88, 0x84, 0x21, 0xA0};
    buffer.insert(buffer.end(), nalu1.begin(), nalu1.end());
 
    // NALU 2: 4 字節(jié)起始碼 + 6 字節(jié)數(shù)據(jù)
    std::vector<uint8_t> nalu2 = {0x00, 0x00, 0x00, 0x01, 0x41, 0x9A, 0x5C, 0xD4, 0x00, 0x11};
    buffer.insert(buffer.end(), nalu2.begin(), nalu2.end());
 
    // NALU 3: 3 字節(jié)起始碼 + 4 字節(jié)數(shù)據(jù)
    std::vector<uint8_t> nalu3 = {0x00, 0x00, 0x01, 0x06, 0x05, 0xFF, 0xEE};
    buffer.insert(buffer.end(), nalu3.begin(), nalu3.end());
 
    // NALU 4: 3 字節(jié)起始碼 + 3 字節(jié)數(shù)據(jù) (測(cè)試末尾)
    std::vector<uint8_t> nalu4 = {0x00, 0x00, 0x01, 0x07, 0xAD, 0xBE};
    buffer.insert(buffer.end(), nalu4.begin(), nalu4.end());
 
    // 輸出構(gòu)建的緩沖區(qū)（可選）
    std::cout << "Constructed H.264 Buffer: ";
    for (size_t i = 0; i < buffer.size(); ++i)
    {
        printf("%02X ", buffer[i]);
    }
    std::cout << "\n\n";
 
    const uint8_t *pnalu = nullptr;
    uint32_t nale_len = 0;
    const uint8_t *pRecord = NULL; // 初始時(shí)為 NULL
 
    int ret;
    int nalu_index = 1;
 
    // 循環(huán)分割并打印每個(gè) NALU
    while ((ret = H264Parse::nalu_tok(buffer.data(), buffer.size(), &pnalu, nale_len, &pRecord)) != -1)
    {
        print_nalu(pnalu, nale_len, nalu_index);
        nalu_index++;
        if (ret == 0)
            break;
    }
 
    if (ret == -1)
    {
        std::cout << "Error occurred during NALU tokenization." << std::endl;
    }
}
 
// 主函數(shù)
int main()
{
    test1();
    // test2();
 
    return 0;
}

以上就是C/C++實(shí)現(xiàn)H264文件解析的詳細(xì)內(nèi)容，更多關(guān)于C++ H264文件解析的資料請(qǐng)關(guān)注腳本之家其它相關(guān)文章！

您可能感興趣的文章: