[Elphel-support] JP4-HDR

[Andrey Filippov]

Andreas,

The JP4* formats are illustrated on
http://community.elphel.com/jp4/jp4demo.php . All of them are targeted
to efficiently compress raw Bayer mosaic data, jp4-hdr does it
somewhat better than jus jp46. But for the de-mosaic algorithm
development there is no difference, this is why I suggested to start
with jp46+ImageJ that is possible to decode with unmodified JPEG
library. That plugin reads (and un-applies) the analog gain settings
together with non-linear gamma conversion with the data provided in
the image metadata. So you just need to adjust the 2 green gains to
different values - the easiest would be to disable automatic white
balance and specify all 4 different gains manually.

Andrey

On Fri, May 28, 2010 at 10:49 PM, Andreas Bean <office at beanbox.com> wrote:
> Hello Andrey,
>
> I'm interested in the example shown at
> http://wiki.elphel.com/index.php?title=JP4_HDR
> The author is Kimstik.
>
> I really want to work on the de-mosaic algorithm processing two
> different greens.
> Are you saying that the two greens are stored in jp4 too and not only in
> jp4-hdr?
>
> Anyway, can you give me your debug version? Because currently I can't
> open the
> jp4 files and thats a problem for me because I can't even start
> developing a
> de-mosaic algorithm.
>
> Andreas
>
>
>
> Andrey Filippov schrieb:
>> Andreas,
>>
>> When I was developing the FPGA encoder for different JP4 flavors I
>> used modified libjpeg - it had a lot of debug output that I needed to
>> troubleshoot encoder. The state of my modification was not really
>> usable for anything but encoder development, I believe Paulo has now
>> more usable code.
>>
>> On the other hand, Jp4-hdr advantage over jp4/ (and even jp46) is just
>> better compression (smaller file size for the same quality), the most
>> challenge of decoding it is not the decoding of the format (you may
>> start with JP46 and ImageJ plugin), but de-mosaic algorithm combined
>> with processing of 2 different greens. And it has to be done in a
>> single step - you need a modified de-mosaic algorithm that is aware of
>> those gains. When that part will work - you can get back to the format
>> and get some extra compression ratio.
>>
>> Andrey
>>
>
>