[FFmpeg-devel] [PATCH v9] lavc/rawdec: Remove monowhite switching code for 1 bpp AVI without a palette
Mats Peterson
matsp888 at yahoo.com
Sat Feb 13 12:52:07 CET 2016
On 02/13/2016 12:50 PM, Mats Peterson wrote:
> On 02/13/2016 12:31 PM, Michael Niedermayer wrote:
>> On Sat, Feb 13, 2016 at 11:44:04AM +0100, Mats Peterson wrote:
>>> On 02/13/2016 11:40 AM, Mats Peterson wrote:
>>>> On 02/13/2016 11:39 AM, Paul B Mahol wrote:
>>>>> On 2/13/16, Mats Peterson <matsp888-at-yahoo.com at ffmpeg.org> wrote:
>>>>>> On 02/13/2016 11:28 AM, Mats Peterson wrote:
>>>>>>>>> Once again, how are we supposed to know how many colors the
>>>>>>>>> palette
>>>>>>>>> consists of at the end of avpkt->data?
>>>>>>>>
>>>>>>>> palette size = packet size - width*height
>>>>>>>>
>>>>>>>
>>>>>>> Nope. You have to take the stride in account as well, not just the
>>>>>>> width
>>>>>>> and height, and it varies in nut, depending on whether you use
>>>>>>> "-vcodec
>>>>>>> copy" or "-vcodec rawvideo".
>>>>>>>
>>>>>>> Mats
>>>>>>>
>>>>>>
>>>>>> And you'll have to calculate the stride by subtracting the palette
>>>>>> size
>>>>> >from the packet size, then divide by the height.
>>>>>
>>>>> Nope, pixels outside of width x height are ignored.
>>>>> _______________________________________________
>>>>
>>>> I need the stride for the alignment to 16 bytes. It's not done for fun.
>>>>
>>>> Mats
>>>>
>>>
>>> And pixels outside of the width are still in the packet, that's
>>> what's called the stride.
>>
>> just double checked, this is not the case in written pal8 nut files
>> nor does the nut spec say anything about that.
>>
>
> Then check an odd-width nut file like the one below. And FFmpeg doesn't
> care about specs when using "-vcodec copy". The stride will be the one
> of the original data, whether it's 2, 4 or some other amount of bytes.
>
> https://drive.google.com/open?id=0B3_pEBoLs0fadkJqREhtOU5lT1E
>
> Mats
>
> _______________________________________________
Sorry, dont' bother about that one. It will have the same width in bytes
as the number of pixels alright. But when using "-vcodec copy", the
situation will be like above.
Mats
More information about the ffmpeg-devel
mailing list