Prev: Lets call it, rating the rubbish
Next: Why Pentax dumped the aging CF card
From: Steve JORDI on 26 Mar 2010 02:56
Hi,
I tried to find an explanation on the web but without success.

Does anybody know why the red color looks so ugly when a digital
picture is saved as JPG?
It looks like it's the only dominant color that gets very pixelated
and grainy.
The cause of this artefact?

Thanks for any clue.

Sincerely,
Steve JORDI
M.Sc. in Geophysics

(Remove the I_REALLY_HATE_SPAMMERS from my email address)
------------------------------------------------
1197 Prangins Email: stevejordiI_REALLY_HATE_SPAMMERS(a)hotmail.com
Switzerland WWW: www.sjordi.com
------------------------------------------------
Volcanoes at www.sjordi.com/volcanoes
------------------------------------------------
From: Robert Spanjaard on 26 Mar 2010 04:13
On Fri, 26 Mar 2010 07:56:33 +0100, Steve JORDI wrote:

> Hi,
> I tried to find an explanation on the web but without success.
>
> Does anybody know why the red color looks so ugly when a digital picture
> is saved as JPG?
> It looks like it's the only dominant color that gets very pixelated and
> grainy.
> The cause of this artefact?

The compression settings.

Both JPEGs in the link below were saved using GIMP with Quality set to 95.
But the upper JPEG was saved with Subsampling set to "2x2, 1x1, 1x1
(smallest file)", while the lower JPEG was saved with "1x1, 1x1, 1x1 (best
quality)".
All three images are displayed at 200% to show the difference. As you can
see, there's very little difference between the original and the high
quality JPEG. The reds (and red containing colors like the orange bars) in
the low quality JPEG are visibly degraded.

http://www.arumes.com/temp/jpegquality.html

--
Regards, Robert http://www.arumes.com
From: Chris Malcolm on 26 Mar 2010 05:58
Steve JORDI <stevejordiI_REALLY_HATE_SPAMMERS(a)hotmail.com> wrote:
> Hi,
> I tried to find an explanation on the web but without success.

> Does anybody know why the red color looks so ugly when a digital
> picture is saved as JPG?
> It looks like it's the only dominant color that gets very pixelated
> and grainy.
> The cause of this artefact?

Because bright red pigments are more easily come by than other
colours, and evolution has tuned plant fruits and flowers and animal
eyes to prefer it and to be most sensitive to small difference in red
quality. So it tends to be both the brightest colour in most images,
which means it's the first to reach saturation and overexposure,
especially when the jpeg quality is set to the vivid oversaturation
most people like, and especially when the colour temperature of the
light is lower than white, which it often is.

All of which then gets exaggerated by jpeg compression artefacts in
the colour to whose quality we are most sensitive. Especially in cheap
cameras which use low quality jpeg compression for reasons of economy.

--
Chris Malcolm
From: Martin Brown on 26 Mar 2010 06:35
Steve JORDI wrote:
> Hi,
> I tried to find an explanation on the web but without success.
>
> Does anybody know why the red color looks so ugly when a digital
> picture is saved as JPG?

Mainly because your eye is strongly tuned to detecting ripe red fruit so
red always stands out much more. The same was noticable on slide films.
The same artefacts happen to saturated blue colours but they are much
rarer in nature. Although not at concerts under coloured spotlights.

Flesh tones are another highly tuned sensitivity. Poor colour indicates
sickness and is best avoided. US NTSC TV newscasters used to drift
between green and purple but now are clamped to unreal pale orange.

> It looks like it's the only dominant color that gets very pixelated
> and grainy.
> The cause of this artefact?

The encoding to YCrCb and then subsampling of the chroma which by
default in JPEG is 2x2, most cameras do 2x1 and you can do 1x1.
Photoshop switches to 1x1 chroma sampling at level 6 in the current
version. Other packages offer an option to switch chroma sampling.

It exploits the limited colour resolution in the eye to save space.
>
> Thanks for any clue.

It isn't really a fault of JPEG so much as a feature of the chroma
subsampling trick used to get some extra compression. Your eye isn't all
that sensitive to colour detail compared to luminance so binning the
chroma 2x2 generally works well and saves a lot of bandwidth. It was
first used on colour TV transmissions.

A demo of the JPEG Chroma subsamping artefacts using red and blue comb
patterns is on my webpage. It is based on pixel level detail in pure red
and blue which is an extreme test case for JPEG reconstruction.

http://www.nezumi.demon.co.uk/photo/jpeg/combtest.htm

A second demo of the generational losses with JPEG using 2x2 chroma
subsampling and 1x1 chroma is at:

http://www.nezumi.demon.co.uk/photo/jpeg/2/jpeg2.htm

If you posted an example of the problem you have encountered it might be
possible to help. There are known issues with commercial JPEG decoders
and monochromatic red and blue images. If that is your problem you are
best off decoding it as pretend monochrome luminance only data.

Regards,
Martin Brown
From: Laurence Payne on 26 Mar 2010 07:12
On 26 Mar 2010 09:58:24 GMT, Chris Malcolm <cam(a)holyrood.ed.ac.uk>
wrote:

>
>Because bright red pigments are more easily come by than other
>colours,

Historically, wasn't blue the original pigment generally available?
 |  Next  |  Last
Pages: 1 2 3 4 5 6 7
Prev: Lets call it, rating the rubbish
Next: Why Pentax dumped the aging CF card