如何将相机输入的图像编码为H.264流

Question

我正在尝试使用Mac OS X 10.9.5上的libx264将MacBook Pro内置FaceTime HD相机的输入图像实时编码为H.264视频流。

以下是我采取的步骤：

1. 使用AVFoundation API(AVCaptureDevice类等)以每秒15帧的速度从相机获取1280x720张32BGRA图像
2. 使用libswscale将图像转换为320x180 YUV420P格式
3. 使用libx264将图像编码为H.264视频流(基线配置文件)

每次从相机获得图像时，我都会应用上述步骤，相信编码器会跟踪编码状态，并在NAL单元可用时生成NAL单元。

由于我想在向编码器提供输入图像的同时获得编码帧，我决定每30帧(2秒)刷新编码器(调用x264_encoder_delayer_frames())。

但是，当我重新启动编码时，编码器会在一段时间后停止(x264_encoder_encode()永远不会返回。)我试着在刷新之前更改帧数，但情况没有改变。

以下是相关代码(我省略了图像捕获代码，因为它看起来没有问题。)

你能指出我可能做错了什么吗？

x264_t *encoder;
x264_param_t param;
// Will be called only first time.
int initEncoder() {
int ret;
if ((ret = x264_param_default_preset(&param, "medium", NULL)) < 0) {
return ret;
}
param.i_csp = X264_CSP_I420;
param.i_width  = 320;
param.i_height = 180;
param.b_vfr_input = 0;
param.b_repeat_headers = 1;
param.b_annexb = 1;
if ((ret = x264_param_apply_profile(&param, "baseline")) < 0) {
return ret;
}
encoder = x264_encoder_open(&param);
if (!encoder) {
return AVERROR_UNKNOWN;
}
return 0;
}
// Will be called from encodeFrame() defined below.
int convertImage(const enum AVPixelFormat srcFmt, const int srcW, const int srcH, const uint8_t *srcData, const enum AVPixelFormat dstFmt, const int dstW, const int dstH, x264_image_t *dstData) {
struct SwsContext *sws_ctx;
int ret;
int src_linesize[4];
uint8_t *src_data[4];
sws_ctx = sws_getContext(srcW, srcH, srcFmt,
dstW, dstH, dstFmt,
SWS_BILINEAR, NULL, NULL, NULL);
if (!sws_ctx) {
return AVERROR_UNKNOWN;
}
if ((ret = av_image_fill_linesizes(src_linesize, srcFmt, srcW)) < 0) {
sws_freeContext(sws_ctx);
return ret;
}
if ((ret = av_image_fill_pointers(src_data, srcFmt, srcH, (uint8_t *) srcData, src_linesize)) < 0) {
sws_freeContext(sws_ctx);
return ret;
}
sws_scale(sws_ctx, (const uint8_t * const*)src_data, src_linesize, 0, srcH, dstData->plane, dstData->i_stride);
sws_freeContext(sws_ctx);
return 0;
}
// Will be called for each frame.
int encodeFrame(const uint8_t *data, const int width, const int height) {
int ret;
x264_picture_t pic;
x264_picture_t pic_out;
x264_nal_t *nal;
int i_nal;
if ((ret = x264_picture_alloc(&pic, param.i_csp, param.i_width, param.i_height)) < 0) {
return ret;
}
if ((ret = convertImage(AV_PIX_FMT_RGB32, width, height, data, AV_PIX_FMT_YUV420P, 320, 180, &pic.img)) < 0) {
x264_picture_clean(&pic);
return ret;
}
if ((ret = x264_encoder_encode(encoder, &nal, &i_nal, &pic, &pic_out)) < 0) {
x264_picture_clean(&pic);
return ret;
}
if(ret) {
for (int i = 0; i < i_nal; i++) {
printNAL(nal + i);
}
}
x264_picture_clean(&pic);
return 0;
}
// Will be called every 30 frames.
int flushEncoder() {
int ret;
x264_nal_t *nal;
int i_nal;
x264_picture_t pic_out;
/* Flush delayed frames */
while (x264_encoder_delayed_frames(encoder)) {
if ((ret = x264_encoder_encode(encoder, &nal, &i_nal, NULL, &pic_out)) < 0) {
return ret;
}
if (ret) {
for (int j = 0; j < i_nal; j++) {
printNAL(nal + j);
}
}
}
}

Answer 1

不应该在每一帧之后刷新延迟帧，而应该在没有更多输入帧时(即在编码结束时)只刷新一次。