jpeg from DSLR workflow: to 16bit or not

[Leonardo Canducci-2]

jpegs from my nikon d50 have 8bit color depth (as any jpeg AFAIK) but
I read that if I plan to edit the histograhm with tools like levels or
curves I'd better use 16bit color depth to avoid posterization, even
when the source is 8bit. Is that true? What's the best way to handle
this tools minimizing quality/information loss when source is a 8bit
jpeg? Should I load the jpeg, then switch to 1bit, then edit (curves,
levels) and the export back to jpeg?

Thanks!
--
Leonardo Canducci
_______________________________________________
Digikam-users mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-users

[Gilles Caulier-4]

This is a wrong workflow. JPEG is always 8 bits encoded...

Use RAW format instead. encoding is > 12 bits... But you need to
postprocess image as your camera do...

Gilles Caulier

2009/9/23 Leonardo Canducci <[hidden email]>:

> jpegs from my nikon d50 have 8bit color depth (as any jpeg AFAIK) but
> I read that if I plan to edit the histograhm with tools like levels or
> curves I'd better use 16bit color depth to avoid posterization, even
> when the source is 8bit. Is that true? What's the best way to handle
> this tools minimizing quality/information loss when source is a 8bit
> jpeg? Should I load the jpeg, then switch to 1bit, then edit (curves,
> levels) and the export back to jpeg?
>
> Thanks!
> --
> Leonardo Canducci
> _______________________________________________
> Digikam-users mailing list
> [hidden email]
> https://mail.kde.org/mailman/listinfo/digikam-users
>

_______________________________________________
Digikam-users mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-users

[Leonardo Canducci-2]

2009/9/23 Gilles Caulier <[hidden email]>:
> This is a wrong workflow. JPEG is always 8 bits encoded...
>
> Use RAW format instead. encoding is > 12 bits... But you need to
> postprocess image as your camera do...
I know jpeg is 8bit but stretching histograms in 16bit color space
should avoid posterization anyway. read the third tip at the end of
this page:
http://www.cambridgeincolour.com/tutorials/posterization.htm

>
> Gilles Caulier
>
> 2009/9/23 Leonardo Canducci <[hidden email]>:
>> jpegs from my nikon d50 have 8bit color depth (as any jpeg AFAIK) but
>> I read that if I plan to edit the histograhm with tools like levels or
>> curves I'd better use 16bit color depth to avoid posterization, even
>> when the source is 8bit. Is that true? What's the best way to handle
>> this tools minimizing quality/information loss when source is a 8bit
>> jpeg? Should I load the jpeg, then switch to 1bit, then edit (curves,
>> levels) and the export back to jpeg?
>>
>> Thanks!
>> --
>> Leonardo Canducci
>> _______________________________________________
>> Digikam-users mailing list
>> [hidden email]
>> https://mail.kde.org/mailman/listinfo/digikam-users
>>
> _______________________________________________
> Digikam-users mailing list
> [hidden email]
> https://mail.kde.org/mailman/listinfo/digikam-users
>

--
Leonardo Canducci
_______________________________________________
Digikam-users mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-users

[Paul Hartman-2]

On Wed, Sep 23, 2009 at 5:58 PM, Leonardo Canducci
<[hidden email]> wrote:
> 2009/9/23 Gilles Caulier <[hidden email]>:
>> This is a wrong workflow. JPEG is always 8 bits encoded...
>>
>> Use RAW format instead. encoding is > 12 bits... But you need to
>> postprocess image as your camera do...
> I know jpeg is 8bit but stretching histograms in 16bit color space
> should avoid posterization anyway. read the third tip at the end of
> this page:
> http://www.cambridgeincolour.com/tutorials/posterization.htm

I think he is saying you should use RAW as source image format on your
camera, not JPG. Then you can avoid the entire issue.
_______________________________________________
Digikam-users mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-users

[Gilles Caulier-4]

2009/9/24 Leonardo Canducci <[hidden email]>:
> 2009/9/23 Gilles Caulier <[hidden email]>:
>> This is a wrong workflow. JPEG is always 8 bits encoded...
>>
>> Use RAW format instead. encoding is > 12 bits... But you need to
>> postprocess image as your camera do...
> I know jpeg is 8bit but stretching histograms in 16bit color space
> should avoid posterization anyway. read the third tip at the end of
> this page:
> http://www.cambridgeincolour.com/tutorials/posterization.htm

This is wrong. Switching 8 to 16 do not feel missing data. Histogram
will be full of holes.

To wrap around this problem you need powerfull algorithm to recompose
histogram. There is nothing like this in opensource world.

Gilles Caulier
_______________________________________________
Digikam-users mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-users

[stefan-119]

>> I know jpeg is 8bit but stretching histograms in 16bit color space
>> should avoid posterization anyway. read the third tip at the end of
>> this page:
>> http://www.cambridgeincolour.com/tutorials/posterization.htm
>
> This is wrong. Switching 8 to 16 do not feel missing data. Histogram
> will be full of holes.
>
> To wrap around this problem you need powerfull algorithm to recompose
> histogram. There is nothing like this in opensource world.
>

Sorry to just jump into that conversation, but I couldn't help to notice
that everyone seems to miss the question.
It's not if the picture itself gains from changing the color depth, but if
filters applied to it do.

Just a little example: Lets say his original image is that of a
checkerboard and it's encoded in just 2 bit - black and white.
Now we change the depth to 8 bit. Of course the image doesnt change. It
still has only two colors, black and white. All the new possible colors are
not used.
No gain whatsoever.

Now we apply a very heavy blur filter to both images. The original 2bit
image will transform into something resembling a black & white noise map.
Very ugly.
But the 8bit image will become a smooth gray picture. So the filter output
is improved by converting the original image to a higher color resolution.


So the answer is: it depends on the filter. If whatever you want to do with
the picture afterwards needs to create new color values for pixels, you
gain from the increased color depth.
But be aware that you need to decrease it again to save the image into .jpg
again. And you may loose anything gained beforhand depending on the
algorithm used for that decrease.

Now someone with more knowledge about digikam needs to say to which filters
this applies.


Best Regards,

Stefan


_______________________________________________
Digikam-users mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-users

[Leonardo Canducci-2]

2009/9/25  <[hidden email]>:

>
>>> I know jpeg is 8bit but stretching histograms in 16bit color space
>>> should avoid posterization anyway. read the third tip at the end of
>>> this page:
>>> http://www.cambridgeincolour.com/tutorials/posterization.htm
>>
>> This is wrong. Switching 8 to 16 do not feel missing data. Histogram
>> will be full of holes.
>>
>> To wrap around this problem you need powerfull algorithm to recompose
>> histogram. There is nothing like this in opensource world.
>>
>
> Sorry to just jump into that conversation, but I couldn't help to notice
> that everyone seems to miss the question.
> It's not if the picture itself gains from changing the color depth, but if
> filters applied to it do.
>
> Just a little example: Lets say his original image is that of a
> checkerboard and it's encoded in just 2 bit - black and white.
> Now we change the depth to 8 bit. Of course the image doesnt change. It
> still has only two colors, black and white. All the new possible colors are
> not used.
> No gain whatsoever.
>
> Now we apply a very heavy blur filter to both images. The original 2bit
> image will transform into something resembling a black & white noise map.
> Very ugly.
> But the 8bit image will become a smooth gray picture. So the filter output
> is improved by converting the original image to a higher color resolution.
>
>
> So the answer is: it depends on the filter. If whatever you want to do with
> the picture afterwards needs to create new color values for pixels, you
> gain from the increased color depth.
> But be aware that you need to decrease it again to save the image into .jpg
> again. And you may loose anything gained beforhand depending on the
> algorithm used for that decrease.
>
> Now someone with more knowledge about digikam needs to say to which filters
> this applies.

That is more or less what I was thinking about. I'm mostly interested
in levels and curves filters because that's what I tweak in pictures
usually (ok, sharpness too sometimes).
Thanks!
--
Leonardo Canducci
_______________________________________________
Digikam-users mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-users