Re: [Libav-user] how to set mjpeg encoding quality programmatically?

[YIRAN LI]

2016-06-09 23:29 GMT+10:00 Carl Eugen Hoyos :

> YIRAN LI  writes:
>
> > AVCodecContext.flags |= CODEC_FLAG_QSCALE
> > and AVCodecContext.global_quality  = qscale * FF_QP2LAMBDA,
> > in my case qscale = 3
>
> This is correct, qscale 3 ensures (with mpeg-4 asp) a
> near-lossless visual quality.
>
>
​Thansk Carl, I think I can use the same way to set quality for mpeg4
video, right?

What't the differenct between it and seting qmin qmax in codeccctx?

Thanks
​


> Carl Eugen
> ___
> Libav-user mailing list
> Libav-user@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/libav-user
>
___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user

Re: [Libav-user] Copying pixels directly from AVFrame->data, possible alignment error.

[Ratin]

On Thu, Jun 9, 2016 at 11:53 AM, Peter Steinbach 
wrote:

> Hi Jepser,
>
> I am not an expert on this (@all please correct what I err upon), but
> AFAIK the data is padded per line! that's why you also have a linesize data
> member of AVFrame:
> https://ffmpeg.org/doxygen/3.0/structAVFrame.html
> So in other words, say you have a 6x6 frame in GRAY8, this would mean that
> AVFrame->data[0] would show for the first 2 lines of the frame:
> item0 item1 item2 item3 item4 item5 unused0 unused1
> item6 item7 item8 item9 item10 item11 unused2 unused3
> ...
> assuming that AVFrame->linesize[0] = 8!
>
> Also, a frame in AV_PIX_FMT_BGR24 encoding would map to
> AVFrame->data[0] being the B channel
> AVFrame->data[1] being the G channel
> AVFrame->data[2] being the R channel
>
> Translating this to your code example, this would give:
>
> uint64_t src_i = (xx+(yy*src_camera_width))*3;
> r = f->pAVFrame->data[2][yy*f->pAVFrame->linesize[2]+xx];
> g = f->pAVFrame->data[1][yy*f->pAVFrame->linesize[1]+xx];
> b = f->pAVFrame->data[0][yy*f->pAVFrame->linesize[0]+xx];
>
> assuming that you f->pAVFrame was already filled by some other function.
> Crossing fingers that I got everything right.
>
> Best,
> P
>
>
Wonder why his frame didn't look all garbled, using only B data and feeding
it to other planes data (assuming those r,g,,b variables are put back into
AVFrame->data) would've messed up the color values so much that would
render into a gardbled frame.



>
> On 09.06.2016 20:32, Jesper Taxbøl wrote:
>
> Sorry about the toppost . Im afraid my phone and me had a slight accident.
>
> What I was trying to add was how I index the the source pixels, hoping it
> can shed light on the offset error im experiencing.
>
> uint64_t src_i = (xx+(yy*src_camera_width))*3;
> r = f->pAVFrame->data[0][src_i+0];
> g = f->pAVFrame->data[0][src_i+1];
> b = f->pAVFrame->data[0][src_i+2];
>
> Kind regards
>
> Jesper
>
>
>
> ___
> Libav-user mailing 
> listLibav-user@ffmpeg.orghttp://ffmpeg.org/mailman/listinfo/libav-user
>
>
>
> ___
> Libav-user mailing list
> Libav-user@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/libav-user
>
>
___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user

Re: [Libav-user] Copying pixels directly from AVFrame->data, possible alignment error.

[Peter Steinbach]

Hi Jepser,

I am not an expert on this (@all please correct what I err upon), but 
AFAIK the data is padded per line! that's why you also have a linesize 
data member of AVFrame:

https://ffmpeg.org/doxygen/3.0/structAVFrame.html
So in other words, say you have a 6x6 frame in GRAY8, this would mean 
that AVFrame->data[0] would show for the first 2 lines of the frame:

item0 item1 item2 item3 item4 item5 unused0 unused1
item6 item7 item8 item9 item10 item11 unused2 unused3
...
assuming that AVFrame->linesize[0] = 8!

Also, a frame in AV_PIX_FMT_BGR24 encoding would map to
AVFrame->data[0] being the B channel
AVFrame->data[1] being the G channel
AVFrame->data[2] being the R channel

Translating this to your code example, this would give:

uint64_t src_i = (xx+(yy*src_camera_width))*3;
r = f->pAVFrame->data[2][yy*f->pAVFrame->linesize[2]+xx];
g = f->pAVFrame->data[1][yy*f->pAVFrame->linesize[1]+xx];
b = f->pAVFrame->data[0][yy*f->pAVFrame->linesize[0]+xx];

assuming that you f->pAVFrame was already filled by some other function.
Crossing fingers that I got everything right.

Best,
P

On 09.06.2016 20:32, Jesper Taxbøl wrote:

Sorry about the toppost . Im afraid my phone and me had a slight accident.

What I was trying to add was how I index the the source pixels, hoping 
it can shed light on the offset error im experiencing.


uint64_t src_i = (xx+(yy*src_camera_width))*3;
r = f->pAVFrame->data[0][src_i+0];
g = f->pAVFrame->data[0][src_i+1];
b = f->pAVFrame->data[0][src_i+2];

Kind regards

Jesper




___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user


___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user

Re: [Libav-user] Copying pixels directly from AVFrame->data, possible alignment error.

[Jesper Taxbøl]

Sorry about the toppost . Im afraid my phone and me had a slight accident.

What I was trying to add was how I index the the source pixels, hoping it
can shed light on the offset error im experiencing.

uint64_t src_i = (xx+(yy*src_camera_width))*3;
r = f->pAVFrame->data[0][src_i+0];
g = f->pAVFrame->data[0][src_i+1];
b = f->pAVFrame->data[0][src_i+2];

Kind regards

Jesper
___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user

Re: [Libav-user] Copying pixels directly from AVFrame->data, possible alignment error.

[Jesper Taxbøl]

uint64_t src_i = (xx+(yy*src_camera_width))*3;

r = f->pAVFrame->data[0][src_i+0];

g = f->pAVFrame->data[0][src_i+1];

b = f->pAVFrame->data[0][src_i+2];
On 9 Jun 2016 3:02 pm, "Jesper Taxbøl"  wrote:

> I am doing a pixel based transformation directly on an AVFrame->data[0]
> array, and I am experiencing the following artefact:
>
> http://i.imgur.com/vnABDjE.png
>
> The source frame I am working on is 3840x2160 pixels in AV_PIX_FMT_BGR24
> format which should place all pixels in the first data array, in BGR order.
>
> The bottom 1/3 part of the picture has a problem with the colors, and I
> suspect that the pAVFrame has an alignment problem and I just by chance hit
> some data that maps to pixels.
>
> Any Idea what I am doing wrong?
>
> Kind regards
>
> Jesper Taxbøl
>
>
___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user

Re: [Libav-user] how to set mjpeg encoding quality programmatically?

[Carl Eugen Hoyos]

YIRAN LI  writes:

> AVCodecContext.flags |= CODEC_FLAG_QSCALE
> and AVCodecContext.global_quality  = qscale * FF_QP2LAMBDA, 
> in my case qscale = 3

This is correct, qscale 3 ensures (with mpeg-4 asp) a 
near-lossless visual quality.

Carl Eugen
___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user

[Libav-user] Copying pixels directly from AVFrame->data, possible alignment error.

[Jesper Taxbøl]

I am doing a pixel based transformation directly on an AVFrame->data[0]
array, and I am experiencing the following artefact:

http://i.imgur.com/vnABDjE.png

The source frame I am working on is 3840x2160 pixels in AV_PIX_FMT_BGR24
format which should place all pixels in the first data array, in BGR order.

The bottom 1/3 part of the picture has a problem with the colors, and I
suspect that the pAVFrame has an alignment problem and I just by chance hit
some data that maps to pixels.

Any Idea what I am doing wrong?

Kind regards

Jesper Taxbøl
___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user

Re: [Libav-user] Where and how to add a custom hardware decoder in libavcodec ?

[ssshukla26]

Can anyone please guide me on this ! 



--
View this message in context: 
http://libav-users.943685.n4.nabble.com/Where-and-how-to-add-a-custom-hardware-decoder-in-libavcodec-tp4662203p4662233.html
Sent from the libav-users mailing list archive at Nabble.com.
___
Libav-user mailing list
Libav-user@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/libav-user