On 25.01.2018 22:43, Carl Eugen Hoyos wrote:
2018-01-25 15:54 GMT+01:00 Tobias Rapp <[email protected]>:
On 25.01.2018 09:25, Tobias Rapp wrote:
Hi,
trying to do the equivalent of FFmpeg command-line output option "-q:v 3"
on library level I came up with the following code (error checking stripped
off):
AVCodecContext *enc_ctx = NULL;
const AVCodec *codec = avcodec_find_encoder_by_name("mjpeg");
int ret;
enc_ctx = avcodec_alloc_context3(codec);
enc_ctx->flags |= AV_CODEC_FLAG_QSCALE;
enc_ctx->global_quality = FF_QP2LAMBDA * 3.0;
ret = avcodec_open2(enc_ctx, enc_ctx->codec, NULL);
Unfortunately it doesn't seem to have any effect on the output image size.
Is there something obvious that I'm missing?
Answering myself: Apparently setting the encoder option is not sufficient
for mjpeg. The encoding quality has to be set for each frame that is passed
to the encoder.
Example:
frame->quality = enc_ctx->global_quality;
frame->pict_type = AV_PICTURE_TYPE_NONE;
ret = avcodec_send_frame(enc_ctx, frame);
Do you know if this was always needed?
I only tested the code with release/3.4 and current master, there it is
necessary to set the quality on each frame. Don't know about older releases.
Also I don't know exactly the purpose behind it. Maybe some filters
allow to adapt frame->quality based on scene-change, etc. But it would
be great if the mjpeg encoder would detect an uninitialized
frame->quality value and set it to enc_ctx->global_quality.
Regards,
Tobias
_______________________________________________
Libav-user mailing list
[email protected]
http://ffmpeg.org/mailman/listinfo/libav-user
