[Libav-user] Help : YUV value for a macroblock in a yuv240p frame

[Paul B Mahol]

On 4/10/19, Boris <ndjoreboris at gmail.com> wrote:
> Le mer. 10 avr. 2019 à 11:57, Paul B Mahol <onemda at gmail.com> a écrit :
>
>> On 4/10/19, Boris <ndjoreboris at gmail.com> wrote:
>> > Le mar. 9 avr. 2019 à 17:44, Simon Brown <simon.k.brown at gmail.com> a
>> écrit
>> > :
>> >
>> >>
>> >>> Thanks,
>> >>> If I understand what you said, I have to proccess like this :
>> >>>
>> >>> //Declaration of yuv arrays
>> >>> uint16_t yval[mb_height * mb_width][256];
>> >>> uint16_t uval[mb_height * mb_width][256];
>> >>> uint16_t vval[mb_height * mb_width][256];
>> >>>
>> >>> for(int mby=0;  mby<mb_height; mby++)
>> >>> {
>> >>>          for(int mbx=0; mbx <mb_width ;mbx++)
>> >>>          {
>> >>>              //Now for this 16x16 macroblock
>> >>>             for(int y=0; y<16;y++)
>> >>>             {
>> >>>                for(int x=0; x<16;x++)
>> >>>                {
>> >>>                  yval[mby*mb_width + mbx][y*16+x] =
>> >>> data[0][linesize[0]*y*16+x]; or yval[mby*mb_width + mbx][y*16+x] =
>> >>> data[0][y*16+x]???
>> >>>                  uval[mby*mb_width + mbx][(y/2)*16+x/2] =
>> >>> data[1][linesize[1]*(y/2)*16+x/2];
>> >>>                  uval[mby*mb_width + mbx][(y/2)*16+x/2] =
>> >>> data[2][linesize[2]*(y/2)*16+x/2];
>> >>>                //Setting of Yuv arrays
>> >>>
>> >>>              }
>> >>> }
>> >>> //Let's go to  an other macroblock
>> >>> }
>> >>> }
>> >>>
>> >>> Regards
>> >>>
>> >>>>
>> >>>> Not quite - the data in data[0] is arranged by pixels, so your
>> >> coordinate into that array must include all the pixels, not just the
>> >> current macroblock.  So you need:
>> >> yval[mby*mb_width + mbx][y*16+x] = data[0][ linesize[0]*(mby*16 + y) +
>> >> (mbx*16 + x) ]
>> >> so you skip all lines up to the one you're processing, including all
>> >> previous macroblocks and all lines already processed in the current
>> >> macroblock using "linesize[0]*(mby*16 + y)" and then all the pixels in
>> >> the
>> >> current line up to the macroblock you're interested in, and any pixels
>> >> already processed in that macroblock with "mbx*16 + x".
>> >>
>> >> I hope that makes sense.
>> >>
>> >> Regards,
>> >> Simon
>> >>
>> >
>> > Hello Simon,
>> >
>> > I do like you tell me but data[0][ linesize[0]*(mby*16 + y) + (mbx*16 +
>> x)
>> > ] gives me some negatives pixel's values.
>> > Is it normal ?
>> > For exemple data[0][ linesize[0]*(mby*16 + y) + (mbx*16 + x) ] =
>> >  -62
>> > -65
>> > -72
>> > My question is to know if it is normal to have negatives pixel's value
>> > (
>> y
>> > = -a; u=-b; v=-c) where a, b and c are integers.
>>
>> Doesn't make sense, post full code.
>>
>
> Hello Paul,
> The code is the following :
> mbx and mby are x,y coordinate (position) of the macroblock in the frame ,
> data is AVFrame frame->data and linesize is AVFrame frame->linesize
> Then :
>  for(int y=0; y<16;y++) //run through the length of the 16x16 macroblock
>             {
>                for(int x=0; x<16;x++) ////run through the width of the
> 16x16 macroblock
>                 {
>  cout<<data[0][ linesize[0]*(mby*16 + y) + (mbx*16 + x) ] <<endl;
>
> }
> }
>
> So I still getting some negatives values. My question is to know if it's
> normal.

It is not normal. Make sure you are not reading past end of data[0].