ffmpeg实现音频resample(重采样)

AVFifoBuffer和音频样本是av_sample_fmt_is_planar的样式采样率讲解，下面上代码

[cpp] view plain copy

1. AVFifoBuffer * m_fifo = NULL;  
2.   
3.   
4.   
5. SwrContext * init_pcm_resample(AVFrame *in_frame, AVFrame *out_frame)  
6. {  
7.     SwrContext * swr_ctx = NULL;  
8.     swr_ctx = swr_alloc();  
9.     if (!swr_ctx)  
10.     {  
11.         printf("swr_alloc error \n");  
12.         return NULL;  
13.     }  
14.     AVCodecContext * audio_dec_ctx = icodec->streams[audio_stream_idx]->codec;  
15.     AVSampleFormat sample_fmt;  
16.     sample_fmt = (AVSampleFormat)m_dwBitsPerSample; //样本  
17.     if (audio_dec_ctx->channel_layout == 0)  
18.     {  
19.         audio_dec_ctx->channel_layout = av_get_default_channel_layout(icodec->streams[audio_stream_idx]->codec->channels);  
20.     }  
21.     /* set options */  
22.     av_opt_set_int(swr_ctx, "in_channel_layout",    audio_dec_ctx->channel_layout, 0);  
23.     av_opt_set_int(swr_ctx, "in_sample_rate",       audio_dec_ctx->sample_rate, 0);  
24.     av_opt_set_sample_fmt(swr_ctx, "in_sample_fmt", audio_dec_ctx->sample_fmt, 0);  
25.   
26.     av_opt_set_int(swr_ctx, "out_channel_layout",    audio_dec_ctx->channel_layout, 0);  
27.     av_opt_set_int(swr_ctx, "out_sample_rate",       audio_dec_ctx->sample_rate, 0);  
28.     av_opt_set_sample_fmt(swr_ctx, "out_sample_fmt", sample_fmt, 0);  
29.     swr_init(swr_ctx);  
30.   
31.     int64_t src_nb_samples = in_frame->nb_samples;  
32.     out_frame->nb_samples = av_rescale_rnd(swr_get_delay(swr_ctx,oaudio_st->codec->sample_rate) + src_nb_samples,  
33.         oaudio_st->codec->sample_rate, oaudio_st->codec->sample_rate, AV_ROUND_UP);  
34.   
35.     int ret = av_samples_alloc(out_frame->data, &out_frame->linesize[0],   
36.         icodec->streams[audio_stream_idx]->codec->channels, out_frame->nb_samples,oaudio_st->codec->sample_fmt,1);  
37.     if (ret codec->channels,  
38.         2048, oaudio_st->codec->sample_fmt, 1);  
39.     m_fifo = av_fifo_alloc(buffersize);  
40.     return swr_ctx;  
41. }  
42.   
43. int preform_pcm_resample(SwrContext * pSwrCtx,AVFrame *in_frame, AVFrame *out_frame)  
44. {  
45.     int ret = 0;  
46.     if (pSwrCtx != NULL)   
47.     {  
48.         ret = swr_convert(pSwrCtx, out_frame->data, out_frame->nb_samples,   
49.             (const uint8_t**)in_frame->data, in_frame->nb_samples);  
50.         if (ret linesize[0], oaudio_st->codec->channels,  
51.             ret, oaudio_st->codec->sample_fmt, 1);  
52.         int sss = av_fifo_size(m_fifo);  
53.         sss = av_fifo_realloc2(m_fifo, av_fifo_size(m_fifo) + out_frame->linesize[0]);  
54.         sss = av_fifo_size(m_fifo);  
55.         av_fifo_generic_write(m_fifo, out_frame->data[0], out_frame->linesize[0], NULL);  
56.   
57.         out_frame->pkt_pts = in_frame->pkt_pts;  
58.         out_frame->pkt_dts = in_frame->pkt_dts;  
59.         //有时pkt_pts和pkt_dts不同，并且pkt_pts是编码前的dts,这里要给avframe传入pkt_dts而不能用pkt_pts  
60.         //out_frame->pts = out_frame->pkt_pts;  
61.         out_frame->pts = in_frame->pkt_dts;  
62.     }  
63.     return 0;  
64. }  
65. void uinit_pcm_resample(AVFrame * poutframe,SwrContext * swr_ctx)  
66. {  
67.     if (poutframe)  
68.     {  
69.         avcodec_free_frame(&poutframe);  
70.         poutframe = NULL;  
71.     }  
72.     if (swr_ctx)  
73.     {  
74.         swr_free(&swr_ctx);  
75.         swr_ctx = NULL;  
76.     }  
77.     //析构pcm分包结构  
78.     if(m_fifo)  
79.     {  
80.         av_fifo_free(m_fifo);  
81.         m_fifo = NULL;  
82.     }  
83. }  
84. int perform_code(int stream_type,AVFrame * picture)  
85. {  
86.     AVCodecContext *cctext = NULL;  
87.     AVPacket pkt_t;  
88.     av_init_packet(&pkt_t);  
89.     pkt_t.data = NULL; // packet data will be allocated by the encoder  
90.     pkt_t.size = 0;  
91.     int frameFinished = 0 ;  
92.   
93.     if (stream_type == AUDIO_ID)  
94.     {  
95.         cctext = oaudio_st->codec;  
96.         //如果进和出的的声道，样本，采样率不同,需要重采样  
97.         if(icodec->streams[audio_stream_idx]->codec->sample_fmt != (AVSampleFormat)m_dwBitsPerSample ||  
98.             icodec->streams[audio_stream_idx]->codec->channels != m_dwChannelCount ||  
99.             icodec->streams[audio_stream_idx]->codec->sample_rate != m_dwFrequency)  
100.         {  
101.             int64_t pts_t = picture->pts;  
102.             int duration_t = (double)cctext->frame_size * (icodec->streams[audio_stream_idx]->time_base.den /icodec->streams[audio_stream_idx]->time_base.num)/   
103.                 icodec->streams[audio_stream_idx]->codec->sample_rate;  
104.   
105.             int frame_bytes = cctext->frame_size * av_get_bytes_per_sample(cctext->sample_fmt)* cctext->channels;  
106.             AVFrame * pFrameResample = avcodec_alloc_frame();  
107.             uint8_t * readbuff = new uint8_t[frame_bytes];  
108.   
109.             if(av_sample_fmt_is_planar(cctext->sample_fmt))  
110.             {  
111.                 frame_bytes /= cctext->channels;  
112.             }  
113.   
114.             while (av_fifo_size(m_fifo) >= frame_bytes) //取出写入的未读的包  
115.             {  
116.                 pFrameResample->nb_samples = cctext->frame_size;  
117.                 av_fifo_generic_read(m_fifo, readbuff, frame_bytes, NULL);  
118.   
119.                 //这里一定要考虑音频分片的问题  
120.                 //如果是分片的avcodec_fill_audio_frame传入的buf是单声道的，但是buf_size 是两个声道加一起的数据量  
121.                 //如果不是分片的avcodec_fill_audio_frame传入的buf是双声道的，buf_size 是两个声道加一起的数据量  
122.                 if(av_sample_fmt_is_planar(cctext->sample_fmt))  
123.                 {  
124.                     avcodec_fill_audio_frame(pFrameResample,cctext->channels,cctext->sample_fmt,readbuff,frame_bytes * cctext->channels,1);  
125.                 }  
126.                 else  
127.                 {                     
128.                     avcodec_fill_audio_frame(pFrameResample,cctext->channels,cctext->sample_fmt,readbuff,frame_bytes,0);  
129.                 }  
130.   
131.                 if(m_is_first_audio_pts == 0)  
132.                 {  
133.                     m_first_audio_pts = pts_t;  
134.                     m_is_first_audio_pts = 1;  
135.                 }  
136.                 pFrameResample->pts = m_first_audio_pts;  
137.                 m_first_audio_pts += duration_t;  
138.   
139.   
140.                 pFrameResample->pts = av_rescale_q_rnd(pFrameResample->pts, icodec->streams[audio_stream_idx]->codec->time_base, oaudio_st->codec->time_base, AV_ROUND_NEAR_INF);  
141.                 nRet = avcodec_encode_audio2(cctext,&pkt_t,pFrameResample,&frameFinished);  
142.                 if (nRet>=0 && frameFinished)  
143.                 {  
144.                     write_frame(ocodec,AUDIO_ID,pkt_t);  
145.                     av_free_packet(&pkt_t);  
146.                 }  
147.             }  
148.             if (readbuff)  
149.             {  
150.                 delete []readbuff;  
151.             }  
152.             if (pFrameResample)  
153.             {  
154.                 av_free(pFrameResample);  
155.                 pFrameResample = NULL;  
156.             }  
157.         }  
158.         else  
159.         {  
160.             nRet = avcodec_encode_audio2(cctext,&pkt_t,picture,&frameFinished);  
161.             if (nRet>=0 && frameFinished)  
162.             {  
163.                 write_frame(ocodec,AUDIO_ID,pkt_t);  
164.                 av_free_packet(&pkt_t);  
165.             }  
166.         }  
167.     }  
168.     else if (stream_type == VIDEO_ID)  
169.     {  
170.         cctext = ovideo_st->codec;  
171.         if(icodec->streams[video_stream_idx]->codec->ticks_per_frame != 1)  
172.         {  
173.             AVRational time_base_video_t;  
174.             time_base_video_t.num = icodec->streams[video_stream_idx]->codec->time_base.num;  
175.             time_base_video_t.den = icodec->streams[video_stream_idx]->codec->time_base.den /icodec->streams[video_stream_idx]->codec->ticks_per_frame;  
176.             picture->pts = av_rescale_q_rnd(picture->pts, time_base_video_t, ovideo_st->codec->time_base, AV_ROUND_NEAR_INF);  
177.         }  
178.         else  
179.         {  
180.             picture->pts = av_rescale_q_rnd(picture->pts, icodec->streams[video_stream_idx]->codec->time_base, ovideo_st->codec->time_base, AV_ROUND_NEAR_INF);  
181.         }  
182.         avcodec_encode_video2(cctext,&pkt_t,picture,&frameFinished);  
183.         picture->pts++;  
184.         if (frameFinished)  
185.         {  
186.             write_frame(ocodec,VIDEO_ID,pkt_t);  
187.             av_free_packet(&pkt_t);  
188.         }  
189.     }  
190.     return 1;  
191. }  

1：由于mp3的sample是1152 aac是1024 有时候将解码的mp3编码成aac时如果不做AVFifoBuffer操作，编码的aac音频sample会比原来的少很多，生成的音频会一卡一卡的明显少声音。

 

2：当要编码的音频样本是av_sample_fmt_is_planar分片的时候需要将解码后的视频添加到AVFrame结构体中：但是如图