YUV to JPG encoding using ffmpeg - c++

Question

I am trying to encode a YVU file and save it as jpg file. but i didn't understand the following
1.why the packet size is size*3.
av_new_packet(&pkt,size*3);`
2.In fread why we using size*3/2.
if(fread(buffer , 1, size*3/2, ptrInputFile)<=0)`
3.how they are filling data here
frame->data[0] = buffer;
frame->data[1] = buffer + siz;
frame->data[2] = buffer + siz*5/4;
code:
AVFormatContext *avFrameContext;
AVOutputFormat *avOutputFormat;
AVStream *avStream;
AVCodecContext *avCodecContext;
AVCodec *avCodec;
AVFrame *frame;
AVPacket pkt;
const char *output = "temp.jpg";
FILE *ptrInputFile;
const char *input = "cuc_view_480x272.yuv";
ptrInputFile = fopen(input ,"rb");
if(!ptrInputFile)
return -1;
avFrameContext = avformat_alloc_context();
avOutputFormat = av_guess_format("mjpeg", NULL, NULL);
if(!avOutputFormat)
return -1;
avFrameContext->oformat = avOutputFormat;
if(avio_open(&avFrameContext->pb ,output ,AVIO_FLAG_READ_WRITE)<0)
return -1;
avStream = avformat_new_stream(avFrameContext,NULL);
if(!avStream)
return -1;
avCodecContext = avStream->codec;
avCodecContext->codec_id = avOutputFormat->video_codec;
avCodecContext->codec_type = AVMEDIA_TYPE_VIDEO;
avCodecContext->pix_fmt = PIX_FMT_YUVJ420P;
avCodecContext->width = 480;
avCodecContext->height = 272;
avCodecContext->time_base.num = 1;
avCodecContext->time_base.den = 25;
avCodec = avcodec_find_encoder(avCodecContext->codec_id);
if(!avCodec)
return -1;
if(avcodec_open2(avCodecContext ,avCodec,NULL)<0)
return -1;
frame = av_frame_alloc();
int size = avpicture_get_size(PIX_FMT_YUVJ420P ,avCodecContext->width, avCodecContext->height);
uint8_t *buffer = (uint8_t*)av_malloc(size*sizeof(uint8_t));
avpicture_fill((AVPicture*)frame, buffer, avCodecContext->pix_fmt ,avCodecContext->width, avCodecContext->height);
//write header
avformat_write_header(avFrameContext, NULL);
int siz = avCodecContext->width*avCodecContext->height;
av_new_packet(&pkt,siz*3);
if(fread(buffer , 1, siz*3/2, ptrInputFile)<=0)
return -1;
frame->data[0] = buffer;
frame->data[1] = buffer + siz;
frame->data[2] = buffer + siz*5/4;

Answer 1

If you look at the format of the yuv420p (wiki) the data is formatted in the file as:
As there are 'siz' length of pixels in the image:
siz length of y value
siz/4 length of u value
siz/4 length of v value
So for question 2: we have siz*3/2 length of data to read.
For question 3: y starts at buffer+0, u starts at buffer+siz, and v starts at buffer+siz*5/4.
As for question 1: I am not sure if the data is converted to RGB. If it is converted then it would require 3 byte for each pixel. Additional code is required to see that.

Answer 2

I don't know much about the code you provided above. But if you are trying to encode yuv video and save as jpeg you can directly use the following command in ffmpeg
ffmpeg -f rawvideo -vcodec rawvideo -s <resolution> -r 25 -pix_fmt yuv420p -i video.yuv -preset ultrafast -qp 0 %d.jpg
replace <resolution> by resolution of your video eg. 1920x1080