[FFmpeg-devel] [PATCH] exr: piz decompression

[Paul B Mahol]

On 10/11/13, Reimar Doeffinger <Reimar.Doeffinger at gmx.de> wrote:
>
>
> On 10.10.2013, at 12:48, Paul B Mahol <onemda at gmail.com> wrote:
>
>> On 10/10/13, Reimar Doeffinger <Reimar.Doeffinger at gmx.de> wrote:
>>> On 09.10.2013, at 15:22, Paul B Mahol <onemda at gmail.com> wrote:
>>>> On 8/11/13, Reimar Doeffinger <Reimar.Doeffinger at gmx.de> wrote:
>>>>> On Sun, Aug 11, 2013 at 04:23:54PM +0200, Piotr Bandurski wrote:
>>>>>> Hi,
>>>>>>
>>>>>>>> Signed-off-by: Paul B Mahol <onemda at gmail.com>
>>>>>>> ---
>>>>>>> libavcodec/exr.c | 477
>>>>>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>>>> 1 file changed, 477 insertions(+)
>>>>>>>
>>>>>>> diff --git a/libavcodec/exr.c b/libavcodec/exr.c
>>>>>>> index 7e9e68c..a65e6ba 100644
>>>>>>> --- a/libavcodec/exr.c
>>>>>>> +++ b/libavcodec/exr.c
>>>>>>
>>>>>> What is the status of this patch?
>>>>>
>>>>> I suspect it is kind of waiting for someone to figure out if this
>>>>> is rather ordinary huffman/VLC coding as I guessed and thus should
>>>>> be using the functions we have for that instead of reimplementing it
>>>>> or if I'm completely off...
>>>>
>>>> I will apply this unless someone shows this really can be used by
>>>> existing
>>>> code.
>>>
>>> What needs to be shown? The generic vlc code should handle basically any
>>> kind of variable length code to symbol conversion.
>>> I think that is what the code does, though I don't fully understand it -
>>> assuming you understand the code that should answer whether it can be
>>> used?
>>> Whether the huffman code could be reused is more tricky, and I don't
>>> really
>>> care about it, it just might have been one way to make more clear what
>>> kind
>>> of concepts this codec uses and how it differs from how it is done
>>> commonly.
>>
>> Looks like neither you nor me can do it so I will apply this as is
>> (with additional
>> issues fixed).
>
> I very much expect I could, though I don't care one bit about this feature,
> so avoiding messy code duplication is the only reason for me to do it.
> Is there a good specification of the format? I definitely don't want to do
> it if I have to reverse-engineer your code into a specification first...

There is no specification for this codec(its is based on PIZ image though),
all documents points to take look at OpenEXR source code.