1. 技术文章
  2. How to get frame by frame from MP4? (MediaCodec)

How to get frame by frame from MP4? (MediaCodec)

由 你。 提交于 2019-11-29 15:40:58

The solution would look something like the ExtractMpegFramesTest, in which MediaCodec is used to generate "external" textures from video frames. In the test code, the frames are rendered to an off-screen pbuffer and then saved as PNG. You would just render them directly.

There are a few problems with this:

  1. MPEG video isn't designed to work well as a random-access database. A common GOP (group of pictures) structure has one "key frame" (essentially a JPEG image) followed by 14 delta frames, which just hold the difference from the previous decoded frame. So if you want frame N, you may have to decode frames N-14 through N-1 first. Not a problem if you're always moving forward (playing a movie onto a texture) or you only store key frames (at which point you've invented a clumsy database of JPEG images).
  2. As mentioned in comments and answers, you're likely to get some visual artifacts. How bad these look depends on the material and your compression rate. Since you're generating the frames, you may be able to reduce this by ensuring that, whenever there's a big change, the first frame is always a key frame.
  3. The firmware that MediaCodec interfaces with may want several frames before it starts producing output, even if you start at a key frame. Seeking around in a stream has a latency cost. See e.g. this post. (Ever wonder why DVRs have smooth fast-forward, but not smooth fast-backward?)
  4. MediaCodec frames passed through SurfaceTexture become "external" textures. These have some limitations vs. normal textures -- performance may be worse, can't use as color buffer in an FBO, etc. If you're just rendering it once per frame at 30fps this shouldn't matter.
  5. MediaMetadataRetriever's getFrameAtTime() method has less-than-desirable performance for the reasons noted above. You're unlikely to get better results by writing it yourself, although you can save a bit of time by skipping the step where it creates a Bitmap object. Also, you passed OPTION_CLOSEST_SYNC in, but that will only produce the results you want if all your frames are sync frames (again, clumsy database of JPEG images). You need to use OPTION_CLOSEST.

If you're just trying to play a movie on a texture (or your problem can be reduced to that), Grafika has some examples. One that may be relevant is TextureFromCamera, which renders the camera video stream on a GLES rect that can be zoomed and rotated. You can replace the camera input with the MP4 playback code from one of the other demos. This'll work fine if you're only playing forward, but if you want to skip around or go backward you'll have trouble.

The problem you're describing sounds pretty similar to what 2D game developers deal with. Doing what they do is probably the best approach.

I can see why it might seem easy to have all your textures in a single file, but this is a really really bad idea.

MP4 is a video codec it is highly optimised for a list of frames which have a high level of similarity to adjacent frames i.e. motion. It is also optimised to be decompressed in sequential order, so using a 'random access' approach will be very inefficient.

To give a bit more detail video codecs store key frames (one a second, but the rate changes) and delta frames the rest of the time. The key frames are independently compressed just like separate images, but the delta frames stored as the difference from one or more other frames. The algorithm assumes this difference will be fairly minimal, after motion compensation has been performed.

So if you want to access a single delta frame you code will have to decompress a nearby key frame and all the delta frames that connect it to the frame you want, this will be much slower than just using single frame JPEG.

In short, use JPEG or PNG to compress your textures and add them all to a single archive file to keep it tidy.

Yes there is way to extract single frames from mp4 video.

In principle, you seem to look for alternative way to load textures, where usual way is GLUtils.texImage2D (which fills texture from a Bitmap).

First, you should consider what others advice, and expect visual artifacts from compression. But assuming that your textures form related textures (e.g. an explosion), getting these from video stream makes sense. For unrelated images you'll get better results using JPG or PNG. And note that mp4 video doesn't have alpha channel, often used in textures.

For the task, you can't use MediaMetadataRetriever, it won't give you needed accuracy to extract all frames.

You'd have to work with MediaCodec and MediaExtractor classes. Android documentation for MediaCodec is detailed.

Actually you'll need to implement kind of customized video player, and add one key function: frame step.

Close thing to this is Android's MediaPlayer, which is complete player, but 1) lacks frame-step, and 2) is rather closed-source because it's implemented by lot of native C++ libraries which are impossible to extend and hard to study.

I advice this with experience of creating a frame-by-frame video player, and I did it by adopting MediaPlayer-Extended, which is written in plain java (no native code), so you can include this in your project and add function that you need. It works with Android's MediaCodec and MediaExtractor.
Somewhere in MediaPlayer class you'd add function for frameStep, and add another signal + function in PlaybackThread to decode just one next frame (in paused mode). However, the implementation of this would be up to you. Result would be that you let decoder to obtain and process single frame, consume the frame, then repeat with next frame. I did it, so I know that this approach works.

Another half of the task is about obtaining the result. A video player (with MediaCodec) outputs frames into a Surface. Your task would be to get the pixels. I know about way how to read RGB bitmap from such surface: you need to create OpenGL Pbuffer EGLSurface, let MediaCodec render into this surface (Android's SurfaceTexture), then read pixels from this surface. This is another nontrivial task, you need to create shader to render EOS texture (the surface), and use GLES20.glReadPixels to obtain RGB pixels into a ByteBuffer. You'd then upload this RGB bitmaps into your textures.
However, as you want to load textures, you may find optimized way how to render the video frame directly into your textures, and avoid moving pixels around.

Hope this helps, and good luck in implementation.

Actually I want to post my implementation for current time.

Here h file

#include <jni.h>
#include <memory>

#include <opencv2/opencv.hpp>

#include "looper.h"
#include "media/NdkMediaCodec.h"
#include "media/NdkMediaExtractor.h"

#ifndef NATIVE_CODEC_NATIVECODECC_H
#define NATIVE_CODEC_NATIVECODECC_H

//Originally took from here https://github.com/googlesamples/android- 
ndk/tree/master/native-codec
//Convert took from here 
https://github.com/kueblert/AndroidMediaCodec/blob/master/nativecodecvideo.cpp

class NativeCodec
{
public:
NativeCodec() = default;

~NativeCodec() = default;

void DecodeDone();

void Pause();

void Resume();

bool createStreamingMediaPlayer(const std::string &filename);

void setPlayingStreamingMediaPlayer(bool isPlaying);

void shutdown();

void rewindStreamingMediaPlayer();

int getFrameWidth() const
{
    return m_frameWidth;
}

int getFrameHeight() const
{
    return m_frameHeight;
}

void getNextFrame(std::vector<unsigned char> &imageData);

private:
struct Workerdata
{
    AMediaExtractor *ex;
    AMediaCodec *codec;
    bool sawInputEOS;
    bool sawOutputEOS;
    bool isPlaying;
    bool renderonce;
};

void Seek();

ssize_t m_bufidx = -1;
int m_frameWidth = -1;
int m_frameHeight = -1;
cv::Size m_frameSize;

Workerdata m_data = {nullptr, nullptr, false, false, false, false};
};

#endif //NATIVE_CODEC_NATIVECODECC_H

Here cc file

#include "native_codec.h"

#include <cassert>
#include "native_codec.h"
#include <jni.h>
#include <cstdio>
#include <cstring>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <cerrno>
#include <climits>
#include "util.h"
#include <android/log.h>
#include <string>
#include <chrono>
#include <android/asset_manager.h>
#include <android/asset_manager_jni.h>

#include <android/log.h>
#include <string>
#include <chrono>

// for native window JNI
#include <android/native_window_jni.h>
#include <android/asset_manager.h>
#include <android/asset_manager_jni.h>

using namespace std;
using namespace std::chrono;

bool NativeCodec::createStreamingMediaPlayer(const std::string &filename)
{
AMediaExtractor *ex = AMediaExtractor_new();
media_status_t err = AMediaExtractor_setDataSource(ex, filename.c_str());;

if (err != AMEDIA_OK)
{
    return false;
}

size_t numtracks = AMediaExtractor_getTrackCount(ex);

AMediaCodec *codec = nullptr;

for (int i = 0; i < numtracks; i++)
{
    AMediaFormat *format = AMediaExtractor_getTrackFormat(ex, i);

    int format_color;

    AMediaFormat_getInt32(format, AMEDIAFORMAT_KEY_COLOR_FORMAT, &format_color);
    bool ok = AMediaFormat_getInt32(format, AMEDIAFORMAT_KEY_WIDTH, &m_frameWidth);
    ok = ok && AMediaFormat_getInt32(format, AMEDIAFORMAT_KEY_HEIGHT, 
 &m_frameHeight);

    if (ok)
    {
        m_frameSize = cv::Size(m_frameWidth, m_frameHeight);
    } else
    {
        //Asking format for frame width / height failed.
    }

    const char *mime;

    if (!AMediaFormat_getString(format, AMEDIAFORMAT_KEY_MIME, &mime))
    {
        return false;
    } else if (!strncmp(mime, "video/", 6))
    {
        // Omitting most error handling for clarity.
        // Production code should check for errors.
        AMediaExtractor_selectTrack(ex, i);
        codec = AMediaCodec_createDecoderByType(mime);
        AMediaCodec_configure(codec, format, nullptr, nullptr, 0);
        m_data.ex = ex;
        m_data.codec = codec;
        m_data.sawInputEOS = false;
        m_data.sawOutputEOS = false;
        m_data.isPlaying = false;
        m_data.renderonce = true;
        AMediaCodec_start(codec);
    }

    AMediaFormat_delete(format);
}

return true;
}

void NativeCodec::getNextFrame(std::vector<unsigned char> &imageData)
{
if (!m_data.sawInputEOS)
{
    m_bufidx = AMediaCodec_dequeueInputBuffer(m_data.codec, 2000);

    if (m_bufidx >= 0)
    {
        size_t bufsize;
        auto buf = AMediaCodec_getInputBuffer(m_data.codec, m_bufidx, &bufsize);
        auto sampleSize = AMediaExtractor_readSampleData(m_data.ex, buf, bufsize);

        if (sampleSize < 0)
        {
            sampleSize = 0;
            m_data.sawInputEOS = true;
        }

        auto presentationTimeUs = AMediaExtractor_getSampleTime(m_data.ex);

        AMediaCodec_queueInputBuffer(m_data.codec, m_bufidx, 0, sampleSize, 
presentationTimeUs,
                                     m_data.sawInputEOS ? 
AMEDIACODEC_BUFFER_FLAG_END_OF_STREAM : 0);

        AMediaExtractor_advance(m_data.ex);
    }
}

if (!m_data.sawOutputEOS)
{
    AMediaCodecBufferInfo info;
    auto status = AMediaCodec_dequeueOutputBuffer(m_data.codec, &info, 0);

    if (status >= 0)
    {
        if (info.flags & AMEDIACODEC_BUFFER_FLAG_END_OF_STREAM)
        {
            __android_log_print(ANDROID_LOG_ERROR, 
 "AMEDIACODEC_BUFFER_FLAG_END_OF_STREAM", "AMEDIACODEC_BUFFER_FLAG_END_OF_STREAM :: %s", 
//
                                "output EOS");

            m_data.sawOutputEOS = true;
        }

        if (info.size > 0)
        {
//                size_t bufsize;
            uint8_t *buf = AMediaCodec_getOutputBuffer(m_data.codec, 
  static_cast<size_t>(status), /*bufsize*/nullptr);
            cv::Mat YUVframe(cv::Size(m_frameSize.width, static_cast<int> 
  (m_frameSize.height * 1.5)), CV_8UC1, buf);

            cv::Mat colImg(m_frameSize, CV_8UC3);
            cv::cvtColor(YUVframe, colImg, CV_YUV420sp2BGR, 3);
            auto dataSize = colImg.rows * colImg.cols * colImg.channels();
            imageData.assign(colImg.data, colImg.data + dataSize);
        }

        AMediaCodec_releaseOutputBuffer(m_data.codec, static_cast<size_t>(status), 
 info.size != 0);

        if (m_data.renderonce)
        {
            m_data.renderonce = false;
            return;
        }
    } else if (status < 0)
    {
        getNextFrame(imageData);
    } else if (status == AMEDIACODEC_INFO_OUTPUT_BUFFERS_CHANGED)
    {
        __android_log_print(ANDROID_LOG_ERROR, 
"AMEDIACODEC_INFO_OUTPUT_BUFFERS_CHANGED", "AMEDIACODEC_INFO_OUTPUT_BUFFERS_CHANGED :: %s", //
                            "output buffers changed");
    } else if (status == AMEDIACODEC_INFO_OUTPUT_FORMAT_CHANGED)
    {
        auto format = AMediaCodec_getOutputFormat(m_data.codec);

        __android_log_print(ANDROID_LOG_ERROR, 
"AMEDIACODEC_INFO_OUTPUT_FORMAT_CHANGED", "AMEDIACODEC_INFO_OUTPUT_FORMAT_CHANGED :: %s", 
 //
                            AMediaFormat_toString(format));

        AMediaFormat_delete(format);
    } else if (status == AMEDIACODEC_INFO_TRY_AGAIN_LATER)
    {
        __android_log_print(ANDROID_LOG_ERROR, "AMEDIACODEC_INFO_TRY_AGAIN_LATER", 
  "AMEDIACODEC_INFO_TRY_AGAIN_LATER :: %s", //
                            "no output buffer right now");
    } else
    {
        __android_log_print(ANDROID_LOG_ERROR, "UNEXPECTED INFO CODE", "UNEXPECTED 
 INFO CODE :: %zd", //
                            status);
    }
}
}

void NativeCodec::DecodeDone()
{
if (m_data.codec != nullptr)
{
    AMediaCodec_stop(m_data.codec);
    AMediaCodec_delete(m_data.codec);
    AMediaExtractor_delete(m_data.ex);
    m_data.sawInputEOS = true;
    m_data.sawOutputEOS = true;
}
}

void NativeCodec::Seek()
{
AMediaExtractor_seekTo(m_data.ex, 0, AMEDIAEXTRACTOR_SEEK_CLOSEST_SYNC);
AMediaCodec_flush(m_data.codec);
m_data.sawInputEOS = false;
m_data.sawOutputEOS = false;

if (!m_data.isPlaying)
{
    m_data.renderonce = true;
}
}

void NativeCodec::Pause()
{
if (m_data.isPlaying)
{
    // flush all outstanding codecbuffer messages with a no-op message
    m_data.isPlaying = false;
}
}

void NativeCodec::Resume()
{
if (!m_data.isPlaying)
{
    m_data.isPlaying = true;
}
}

void NativeCodec::setPlayingStreamingMediaPlayer(bool isPlaying)
{
if (isPlaying)
{
    Resume();
} else
{
    Pause();
}
}

void NativeCodec::shutdown()
{
m_bufidx = -1;
DecodeDone();
}

void NativeCodec::rewindStreamingMediaPlayer()
{
Seek();
}

So, according to this implementation for format conversion (in my case from YUV to BGR) you need to set up OpenCV, for understand how to do it check this two source

https://www.youtube.com/watch?v=jN9Bv5LHXMk

https://www.youtube.com/watch?v=0fdIiOqCz3o

And also for sample I leave here my CMakeLists.txt file

#For add OpenCV take a look at this video
#https://www.youtube.com/watch?v=jN9Bv5LHXMk
#https://www.youtube.com/watch?v=0fdIiOqCz3o
#Look at the video than compare with this file and make the same

set(pathToProject
    C:/Users/tetavi/Downloads/Buffer/OneMoreArNew/arcore-android- 
sdk/samples/hello_ar_c)
set(pathToOpenCv C:/OpenCV-android-sdk)

cmake_minimum_required(VERSION 3.4.1)

set(CMAKE VERBOSE MAKEFILE on)
set(CMAKE CXX FLAGS "${CMAKE_CXX_FLAGS} -std=gnu++11")

include_directories(${pathToOpenCv}/sdk/native/jni/include)

# Import the ARCore library.
add_library(arcore SHARED IMPORTED)
set_target_properties(arcore PROPERTIES IMPORTED_LOCATION
    ${ARCORE_LIBPATH}/${ANDROID_ABI}/libarcore_sdk_c.so
    INTERFACE_INCLUDE_DIRECTORIES ${ARCORE_INCLUDE}
    )

# Import the glm header file from the NDK.
add_library(glm INTERFACE)
set_target_properties(glm PROPERTIES
    INTERFACE_INCLUDE_DIRECTORIES 
${ANDROID_NDK}/sources/third_party/vulkan/src/libs/glm
    )

# This is the main app library.
add_library(hello_ar_native SHARED
     src/main/cpp/background_renderer.cc
    src/main/cpp/hello_ar_application.cc
    src/main/cpp/jni_interface.cc
    src/main/cpp/video_render.cc
    src/main/cpp/geometry_loader.cc
    src/main/cpp/plane_renderer.cc
    src/main/cpp/native_codec.cc
    src/main/cpp/point_cloud_renderer.cc
    src/main/cpp/frame_manager.cc
    src/main/cpp/safe_queue.cc
    src/main/cpp/stb_image.h
    src/main/cpp/util.cc)

add_library(lib_opencv SHARED IMPORTED)
set_target_properties(lib_opencv PROPERTIES IMPORTED_LOCATION

${pathToProject}/app/src/main/jniLibs/${CMAKE_ANDROID_ARCH_ABI}/libopencv_java3.so)

target_include_directories(hello_ar_native PRIVATE
    src/main/cpp)

target_link_libraries(hello_ar_native $\{log-lib} lib_opencv
    android
    log
    GLESv2
    glm
    mediandk
    arcore)

Usage:

You need to create stream media player with this method 

NaviteCodec::createStreamingMediaPlayer(pathToYourMP4file);

and then just use 

NativeCodec::getNextFrame(imageData);

Feel free to ask

标签
android
opengl-es
mp4
mediacodec
android-mediacodec
易学教程内所有资源均来自网络或用户发布的内容，如有违反法律规定的内容欢迎反馈！
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!