当前位置： 首页 > news >正文

Android平台FFmpeg视频解码全流程指南

news 来源：原创 2025/5/9 7:34:37

本文将详细介绍在Android平台上使用FFmpeg进行高效视频解码的实现方案，采用面向对象的设计思想。

一、架构设计

1.1 整体架构
采用三层架构设计：
• 应用层：提供用户接口和UI展示

• 业务逻辑层：管理解码流程和状态

• Native层：FFmpeg核心解码实现

1.2 状态管理方案
使用静态常量替代枚举类：

public class DecodeState {public static final int STATE_IDLE = 0;public static final int STATE_PREPARING = 1;public static final int STATE_READY = 2;public static final int STATE_DECODING = 3;public static final int STATE_PAUSED = 4;public static final int STATE_STOPPED = 5;public static final int STATE_ERROR = 6;
}

二、核心类实现

2.1 视频帧数据封装类

public class VideoFrame {private final byte[] videoData;private final int width;private final int height;private final long pts;private final int format;private final int rotation;public VideoFrame(byte[] videoData, int width, int height, long pts, int format, int rotation) {this.videoData = videoData;this.width = width;this.height = height;this.pts = pts;this.format = format;this.rotation = rotation;}// Getter方法public byte[] getVideoData() {return videoData;}public int getWidth() {return width;}public int getHeight() {return height;}public long getPts() {return pts;}public int getFormat() {return format;}public int getRotation() {return rotation;}// 转换为Bitmappublic Bitmap toBitmap() {YuvImage yuvImage = new YuvImage(videoData, ImageFormat.NV21, width, height, null);ByteArrayOutputStream os = new ByteArrayOutputStream();yuvImage.compressToJpeg(new Rect(0, 0, width, height), 100, os);byte[] jpegByteArray = os.toByteArray();Bitmap bitmap = BitmapFactory.decodeByteArray(jpegByteArray, 0, jpegByteArray.length);// 处理旋转if (rotation != 0) {Matrix matrix = new Matrix();matrix.postRotate(rotation);bitmap = Bitmap.createBitmap(bitmap, 0, 0, bitmap.getWidth(), bitmap.getHeight(), matrix, true);}return bitmap;}
}

2.2 视频解码器封装类

public class VideoDecoder {// 解码状态常量public static final int STATE_IDLE = 0;public static final int STATE_PREPARING = 1;public static final int STATE_READY = 2;public static final int STATE_DECODING = 3;public static final int STATE_PAUSED = 4;public static final int STATE_STOPPED = 5;public static final int STATE_ERROR = 6;// 错误码常量public static final int ERROR_CODE_FILE_NOT_FOUND = 1001;public static final int ERROR_CODE_UNSUPPORTED_FORMAT = 1002;public static final int ERROR_CODE_DECODE_FAILED = 1003;private volatile int currentState = STATE_IDLE;private long nativeHandle;private Handler mainHandler;public interface DecodeListener {void onFrameDecoded(VideoFrame frame);void onDecodeFinished();void onErrorOccurred(int errorCode, String message);void onStateChanged(int newState);}public VideoDecoder() {nativeHandle = nativeInit();mainHandler = new Handler(Looper.getMainLooper());}public void prepare(String filePath) {if (currentState != STATE_IDLE) {notifyError(ERROR_CODE_DECODE_FAILED, "Decoder is not in idle state");return;}setState(STATE_PREPARING);new Thread(() -> {boolean success = nativePrepare(nativeHandle, filePath);if (success) {setState(STATE_READY);} else {setState(STATE_ERROR);notifyError(ERROR_CODE_FILE_NOT_FOUND, "Failed to prepare decoder");}}).start();}public void startDecoding(DecodeListener listener) {if (currentState != STATE_READY && currentState != STATE_PAUSED) {notifyError(ERROR_CODE_DECODE_FAILED, "Decoder is not ready");return;}setState(STATE_DECODING);new Thread(() -> {nativeStartDecoding(nativeHandle, listener);setState(STATE_STOPPED);}).start();}public void pause() {if (currentState == STATE_DECODING) {setState(STATE_PAUSED);nativePause(nativeHandle);}}public void resume() {if (currentState == STATE_PAUSED) {setState(STATE_DECODING);nativeResume(nativeHandle);}}public void stop() {setState(STATE_STOPPED);nativeStop(nativeHandle);}public void release() {setState(STATE_STOPPED);nativeRelease(nativeHandle);nativeHandle = 0;}public int getCurrentState() {return currentState;}private void setState(int newState) {currentState = newState;mainHandler.post(() -> {if (listener != null) {listener.onStateChanged(newState);}});}private void notifyError(int errorCode, String message) {mainHandler.post(() -> {if (listener != null) {listener.onErrorOccurred(errorCode, message);}});}// Native方法private native long nativeInit();private native boolean nativePrepare(long handle, String filePath);private native void nativeStartDecoding(long handle, DecodeListener listener);private native void nativePause(long handle);private native void nativeResume(long handle);private native void nativeStop(long handle);private native void nativeRelease(long handle);static {System.loadLibrary("avcodec");System.loadLibrary("avformat");System.loadLibrary("avutil");System.loadLibrary("swscale");System.loadLibrary("ffmpeg-wrapper");}
}

三、Native层实现

3.1 上下文结构体

typedef struct {AVFormatContext *format_ctx;AVCodecContext *codec_ctx;int video_stream_idx;SwsContext *sws_ctx;volatile int is_decoding;volatile int is_paused;int video_width;int video_height;int rotation;
} VideoDecodeContext;

3.2 JNI接口实现

// 初始化解码器
JNIEXPORT jlong JNICALL
Java_com_example_VideoDecoder_nativeInit(JNIEnv *env, jobject thiz) {VideoDecodeContext *ctx = (VideoDecodeContext *)malloc(sizeof(VideoDecodeContext));memset(ctx, 0, sizeof(VideoDecodeContext));ctx->is_decoding = 0;ctx->is_paused = 0;ctx->rotation = 0;return (jlong)ctx;
}// 准备解码器
JNIEXPORT jboolean JNICALL
Java_com_example_VideoDecoder_nativePrepare(JNIEnv *env, jobject thiz, jlong handle, jstring file_path) {VideoDecodeContext *ctx = (VideoDecodeContext *)handle;const char *path = (*env)->GetStringUTFChars(env, file_path, NULL);// 打开媒体文件if (avformat_open_input(&ctx->format_ctx, path, NULL, NULL) != 0) {LOGE("Could not open file: %s", path);(*env)->ReleaseStringUTFChars(env, file_path, path);return JNI_FALSE;}// 获取流信息if (avformat_find_stream_info(ctx->format_ctx, NULL) < 0) {LOGE("Could not find stream information");(*env)->ReleaseStringUTFChars(env, file_path, path);avformat_close_input(&ctx->format_ctx);return JNI_FALSE;}// 查找视频流ctx->video_stream_idx = -1;for (int i = 0; i < ctx->format_ctx->nb_streams; i++) {if (ctx->format_ctx->streams[i]->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {ctx->video_stream_idx = i;// 获取视频旋转信息AVDictionaryEntry *rotate_tag = av_dict_get(ctx->format_ctx->streams[i]->metadata, "rotate", NULL, 0);if (rotate_tag && rotate_tag->value) {ctx->rotation = atoi(rotate_tag->value);}break;}}// 检查是否找到视频流if (ctx->video_stream_idx == -1) {LOGE("Could not find video stream");(*env)->ReleaseStringUTFChars(env, file_path, path);avformat_close_input(&ctx->format_ctx);return JNI_FALSE;}// 获取解码器参数AVCodecParameters *codec_params = ctx->format_ctx->streams[ctx->video_stream_idx]->codecpar;AVCodec *decoder = avcodec_find_decoder(codec_params->codec_id);if (!decoder) {LOGE("Unsupported codec");(*env)->ReleaseStringUTFChars(env, file_path, path);avformat_close_input(&ctx->format_ctx);return JNI_FALSE;}// 创建解码上下文ctx->codec_ctx = avcodec_alloc_context3(decoder);avcodec_parameters_to_context(ctx->codec_ctx, codec_params);// 打开解码器if (avcodec_open2(ctx->codec_ctx, decoder, NULL) < 0) {LOGE("Could not open codec");(*env)->ReleaseStringUTFChars(env, file_path, path);avcodec_free_context(&ctx->codec_ctx);avformat_close_input(&ctx->format_ctx);return JNI_FALSE;}// 保存视频尺寸ctx->video_width = ctx->codec_ctx->width;ctx->video_height = ctx->codec_ctx->height;(*env)->ReleaseStringUTFChars(env, file_path, path);return JNI_TRUE;
}

3.3 核心解码逻辑

// 开始解码
JNIEXPORT void JNICALL
Java_com_example_VideoDecoder_nativeStartDecoding(JNIEnv *env, jobject thiz, jlong handle, jobject listener) {VideoDecodeContext *ctx = (VideoDecodeContext *)handle;ctx->is_decoding = 1;ctx->is_paused = 0;// 获取Java回调方法和类jclass listener_class = (*env)->GetObjectClass(env, listener);jmethodID on_frame_method = (*env)->GetMethodID(env, listener_class, "onFrameDecoded", "(Lcom/example/VideoFrame;)V");jmethodID on_finish_method = (*env)->GetMethodID(env, listener_class, "onDecodeFinished", "()V");jmethodID on_error_method = (*env)->GetMethodID(env, listener_class, "onErrorOccurred", "(ILjava/lang/String;)V");// 分配帧和包AVFrame *frame = av_frame_alloc();AVFrame *rgb_frame = av_frame_alloc();AVPacket *packet = av_packet_alloc();// 准备图像转换上下文 (转换为RGB24)ctx->sws_ctx = sws_getContext(ctx->video_width, ctx->video_height, ctx->codec_ctx->pix_fmt,ctx->video_width, ctx->video_height, AV_PIX_FMT_RGB24,SWS_BILINEAR, NULL, NULL, NULL);if (!ctx->sws_ctx) {(*env)->CallVoidMethod(env, listener, on_error_method, VideoDecoder.ERROR_CODE_DECODE_FAILED,(*env)->NewStringUTF(env, "Could not initialize sws context"));goto end;}// 分配RGB缓冲区int rgb_buffer_size = av_image_get_buffer_size(AV_PIX_FMT_RGB24, ctx->video_width, ctx->video_height, 1);uint8_t *rgb_buffer = (uint8_t *)av_malloc(rgb_buffer_size);av_image_fill_arrays(rgb_frame->data, rgb_frame->linesize, rgb_buffer,AV_PIX_FMT_RGB24, ctx->video_width, ctx->video_height, 1);// 解码循环while (ctx->is_decoding && av_read_frame(ctx->format_ctx, packet) >= 0) {if (packet->stream_index == ctx->video_stream_idx) {// 发送到解码器if (avcodec_send_packet(ctx->codec_ctx, packet) == 0) {// 接收解码后的帧while (avcodec_receive_frame(ctx->codec_ctx, frame) == 0) {if (!ctx->is_decoding) break;// 等待暂停状态结束while (ctx->is_paused && ctx->is_decoding) {usleep(10000); // 10ms}if (!ctx->is_decoding) break;// 转换像素格式sws_scale(ctx->sws_ctx, (const uint8_t *const *)frame->data,frame->linesize, 0, ctx->video_height,rgb_frame->data, rgb_frame->linesize);// 创建Java VideoFrame对象jclass frame_class = (*env)->FindClass(env, "com/example/VideoFrame");jmethodID frame_ctor = (*env)->GetMethodID(env, frame_class, "<init>", "([BIIJI)V");// 创建字节数组jbyteArray rgb_array = (*env)->NewByteArray(env, rgb_buffer_size);(*env)->SetByteArrayRegion(env, rgb_array, 0, rgb_buffer_size, (jbyte *)rgb_buffer);// 创建VideoFrame对象jobject video_frame = (*env)->NewObject(env, frame_class, frame_ctor,rgb_array, ctx->video_width,ctx->video_height,frame->pts,AV_PIX_FMT_RGB24,ctx->rotation);// 回调到Java层(*env)->CallVoidMethod(env, listener, on_frame_method, video_frame);// 释放本地引用(*env)->DeleteLocalRef(env, video_frame);(*env)->DeleteLocalRef(env, rgb_array);}}}av_packet_unref(packet);}// 解码完成回调if (ctx->is_decoding) {(*env)->CallVoidMethod(env, listener, on_finish_method);}end:// 释放资源if (rgb_buffer) av_free(rgb_buffer);if (ctx->sws_ctx) sws_freeContext(ctx->sws_ctx);av_frame_free(&frame);av_frame_free(&rgb_frame);av_packet_free(&packet);
}

四、使用示例

public class VideoPlayerActivity extends AppCompatActivity implements VideoDecoder.DecodeListener {private VideoDecoder videoDecoder;private ImageView videoView;private Button btnPlay, btnPause, btnStop;@Overrideprotected void onCreate(Bundle savedInstanceState) {super.onCreate(savedInstanceState);setContentView(R.layout.activity_video_player);videoView = findViewById(R.id.video_view);btnPlay = findViewById(R.id.btn_play);btnPause = findViewById(R.id.btn_pause);btnStop = findViewById(R.id.btn_stop);videoDecoder = new VideoDecoder();// 准备视频文件String videoPath = getExternalFilesDir(null) + "/test.mp4";// 设置按钮点击监听btnPlay.setOnClickListener(v -> {if (videoDecoder.getCurrentState() == VideoDecoder.STATE_READY || videoDecoder.getCurrentState() == VideoDecoder.STATE_PAUSED) {videoDecoder.startDecoding(this);} else if (videoDecoder.getCurrentState() == VideoDecoder.STATE_IDLE) {videoDecoder.prepare(videoPath);}});btnPause.setOnClickListener(v -> {if (videoDecoder.getCurrentState() == VideoDecoder.STATE_DECODING) {videoDecoder.pause();}});btnStop.setOnClickListener(v -> {if (videoDecoder.getCurrentState() != VideoDecoder.STATE_IDLE && videoDecoder.getCurrentState() != VideoDecoder.STATE_STOPPED) {videoDecoder.stop();}});}@Overridepublic void onFrameDecoded(VideoFrame frame) {runOnUiThread(() -> {Bitmap bitmap = frame.toBitmap();videoView.setImageBitmap(bitmap);});}@Overridepublic void onDecodeFinished() {runOnUiThread(() -> {Toast.makeText(this, "解码完成", Toast.LENGTH_SHORT).show();videoView.setImageBitmap(null);});}@Overridepublic void onErrorOccurred(int errorCode, String message) {runOnUiThread(() -> {String errorMsg = "错误(" + errorCode + "): " + message;Toast.makeText(this, errorMsg, Toast.LENGTH_LONG).show();});}@Overridepublic void onStateChanged(int newState) {runOnUiThread(() -> updateUI(newState));}private void updateUI(int state) {btnPlay.setEnabled(state == VideoDecoder.STATE_READY || state == VideoDecoder.STATE_PAUSED ||state == VideoDecoder.STATE_IDLE);btnPause.setEnabled(state == VideoDecoder.STATE_DECODING);btnStop.setEnabled(state == VideoDecoder.STATE_DECODING || state == VideoDecoder.STATE_PAUSED);}@Overrideprotected void onDestroy() {super.onDestroy();videoDecoder.release();}
}

五、性能优化建议

  1. 使用Surface直接渲染：
    • 通过ANativeWindow直接渲染YUV数据，避免格式转换

    • 减少内存拷贝和Bitmap创建开销

  2. 硬解码优先：

    // 在nativePrepare中检测硬件解码器
AVCodec *decoder = NULL;
if (isHardwareDecodeSupported(codec_id)) {decoder = avcodec_find_decoder_by_name("h264_mediacodec");
}
if (!decoder) {decoder = avcodec_find_decoder(codec_id);
}

  3. 帧缓冲队列优化：
    • 实现生产者-消费者模型

    • 设置合理的队列大小(3-5帧)

    • 丢帧策略处理视频不同步问题

  4. 多线程处理：
    • 分离解码线程和渲染线程

    • 使用线程池处理耗时操作

  5. 内存复用：

    // 复用AVPacket和AVFrame
static AVPacket *reuse_packet = NULL;
if (!reuse_packet) {reuse_packet = av_packet_alloc();
} else {av_packet_unref(reuse_packet);
}

  6. 精准帧率控制：

    // 根据帧率控制解码速度
AVRational frame_rate = ctx->format_ctx->streams[ctx->video_stream_idx]->avg_frame_rate;
double frame_delay = av_q2d(av_inv_q(frame_rate)) * 1000000; // 微秒int64_t last_frame_time = av_gettime();
while (decoding) {// ...解码逻辑...int64_t current_time = av_gettime();int64_t elapsed = current_time - last_frame_time;if (elapsed < frame_delay) {usleep(frame_delay - elapsed);}last_frame_time = av_gettime();
}

  7. 低功耗优化：
    • 根据设备温度调整解码策略

    • 在后台时降低帧率或暂停解码

六、兼容性处理

  1. API版本适配：

    private static boolean isSurfaceTextureSupported() {return Build.VERSION.SDK_INT >= Build.VERSION_CODES.ICE_CREAM_SANDWICH;
}

  2. 权限处理：

    private boolean checkStoragePermission() {if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.M) {return checkSelfPermission(Manifest.permission.READ_EXTERNAL_STORAGE) == PackageManager.PERMISSION_GRANTED;}return true;
}

  3. ABI兼容：

    android {defaultConfig {ndk {abiFilters 'armeabi-v7a', 'arm64-v8a', 'x86', 'x86_64'}}
}

七、错误处理与日志

  1. 完善的错误处理：

    public void onErrorOccurred(int errorCode, String message) {switch (errorCode) {case VideoDecoder.ERROR_CODE_FILE_NOT_FOUND:// 处理文件不存在错误break;case VideoDecoder.ERROR_CODE_UNSUPPORTED_FORMAT:// 处理不支持的格式错误break;default:// 处理未知错误}
}

  2. 日志系统：

    #define LOG_LEVEL_VERBOSE 1
#define LOG_LEVEL_DEBUG   2
#define LOG_LEVEL_INFO    3
#define LOG_LEVEL_WARN    4
#define LOG_LEVEL_ERROR   5void log_print(int level, const char *tag, const char *fmt, ...) {if (level >= CURRENT_LOG_LEVEL) {va_list args;va_start(args, fmt);__android_log_vprint(level, tag, fmt, args);va_end(args);}
}

八、扩展功能

  1. 视频信息获取：

    public class VideoInfo {public int width;public int height;public long duration;public float frameRate;public int rotation;
}// 在VideoDecoder中添加方法
public VideoInfo getVideoInfo() {return nativeGetVideoInfo(nativeHandle);
}

  2. 视频截图功能：

    public Bitmap captureFrame() {if (currentState == STATE_DECODING || currentState == STATE_PAUSED) {return nativeCaptureFrame(nativeHandle);}return null;
}

  3. 视频缩放控制：

    // 在native层实现缩放
sws_scale(ctx->sws_ctx, frame->data, frame->linesize, 0, ctx->video_height, scaled_frame->data, scaled_frame->linesize);

九、测试建议

  1. 单元测试：

    @Test
public void testDecoderStates() {VideoDecoder decoder = new VideoDecoder();assertEquals(VideoDecoder.STATE_IDLE, decoder.getCurrentState());decoder.prepare("test.mp4");// 等待准备完成assertEquals(VideoDecoder.STATE_READY, decoder.getCurrentState());
}

  2. 性能测试：

    long startTime = System.currentTimeMillis();
// 执行解码操作
long endTime = System.currentTimeMillis();
Log.d("Performance", "解码耗时: " + (endTime - startTime) + "ms");

  3. 内存泄漏检测：
    • 使用Android Profiler监控内存使用

    • 重复创建释放解码器检查内存增长

十、总结

本文实现的Android FFmpeg视频解码方案具有以下特点：

  1. 高性能：通过Native层优化和合理的内存管理实现高效解码
  2. 高兼容性：避免使用枚举类，支持广泛的Android设备
  3. 可扩展性：模块化设计便于添加新功能
  4. 稳定性：完善的状态管理和错误处理机制
  5. 易用性：清晰的API接口和完整的文档

开发者可以根据实际需求在此基础框架上进行扩展，如添加音频解码、视频滤镜等功能，构建更完整的媒体播放解决方案。

相关文章：

  • 跨平台移动开发框架React Native和Flutter性能对比
  • GuPPy-v1.2.0安装与使用-生信工具52
  • 数字孪生医疗：构建患者特异性数字孪生体路径探析
  • JVM运行时数据区域(Run-Time Data Areas)的解析
  • 关于 wordpress 统计访问量初始数值错误的解决方法
  • Qt获取CPU使用率及内存占用大小
  • typecho中的Widget设计文档
  • 17.thinkphp的分页功能
  • 广州AI数字人：从“虚拟”走向“现实”的变革力量
  • 软件工程（五）：设计模式
  • 体绘制中的传输函数（transfer func）介绍
  • 网站公安备案流程及审核时间
  • Django进阶：用户认证、REST API与Celery异步任务全解析
  • flutter build apk出现的一些奇怪的编译错误
  • 探索 C++ 在行业应用与技术融合中的核心价值
  • 前端面试宝典---JavaScript import 与 Node.js require 的区别
  • 华为HCIP-AI认证考试版本更新通知
  • Open CASCADE学习|Geom2d_Curve类
  • nginx 实现动静分离
  • OpenCv实战笔记（3）基于opencv实现调用摄像头并实时显示画面
  • 习近平同俄罗斯总统普京会谈
  • 国家主席习近平同普京总统共见记者
  • “80后”计算机专家唐金辉已任南京林业大学副校长
  • 深入贯彻中央八项规定精神学习教育中央第一指导组指导督导河北省见面会召开
  • 李云泽：支持小微企业、民企融资一揽子政策将从增供给、降成本、提效率、优环境4个方面发力
  • 外交部：中方和欧洲议会决定同步全面取消对相互交往的限制