Explanation Of Graininess Please.

For those who have some experience with camera picture processing, what is it
that causes differences in the portions of a picture much further than what
would be expected?

The following example is a picture taken of a house across the street, using
a Canon Digital Rebel XS, at ISO 100. The photo was taken out-of-focus on
purpose, to demonstrate the capture of the sensor. It was done to highlight
the graininess seen in pictures when blown up to the pixel level.

What is the reason that this capture shows multiple artifacts? Is it Canon's
particular in-camera RAW capture that does this?


I don't care about the JPG ones as the artefacts also show up in the RAW
file, even before the JPEG conversion. Also never mind the dead-pixel that
shows in the lower right corner of the picture.



http://i172.photobucket.com/albums/w..._1918rt.jpg?t=
1299705139

or just

http://tinyurl.com/4t8b76b


It doesn't seem to matter which RAW viewer is used. I viewed the picture
with both Canon's and RawTherapee's. The posted pic came from the RT capture
to JPEG conversion. I don't know if the site allows uploads larger than
10Megs.

I'd swear that taking a picture without any lens on it would result in the
same capture. Explanations?


--
__
SneakyP
To email me, you know what to do.

Supernews, if you get a complaint from a Jamie Baillie, please see:
http://www.canadianisp.ca/jamie_baillie.html

otter wrote in news:a3088e80-dd24-425f-a579-
:

OK, good, I'm not the only one. I don't see anything other than a
picture that is very out of focus.



"CTRL +" the picture (on Firefox browser) until it's obvious the view of a
part of the picture is showing a "mottled" appearance, rather than a smooth
one. It is more visible in the green areas.

A mild gaussian blur would demonstrate the removal of such patterns.

An unsharp mask would reveal the granulation even more.


Is this "mottling" what is called luminance noise?



Oh, and to the poster who spat out an obnoxious diatribe about killfiltering,
blow it out your ass, troll.




--
__
SneakyP
To email me, you know what to do.

Supernews, if you get a complaint from a Jamie Baillie, please see:
http://www.canadianisp.ca/jamie_baillie.html

On Wed, 9 Mar 2011 19:45:33 -0800 (PST), otter
wrote:

On Mar 9, 9:38*pm, shiva das wrote:
In article ,



*SneakyP wrote:
For those who have some experience with camera picture processing, what is it
that causes differences in the portions of a picture much further than what
would be expected? *

The following example is a picture taken of a house across the street, using
a Canon Digital Rebel XS, at ISO 100. *The photo was taken out-of-focus on
purpose, to demonstrate the capture of the sensor. *It was done to highlight
the graininess seen in pictures when blown up to the pixel level.

*What is the reason that this capture shows multiple artifacts? Is it Canon's
particular in-camera RAW capture that does this?

I don't care about the JPG ones as the artefacts also show up in the RAW
file, even before the JPEG conversion. *Also never mind the dead-pixel that
shows in the lower right corner of the picture.

http://i172.photobucket.com/albums/w..._09_IMG_1918rt...
1299705139

or just

http://tinyurl.com/4t8b76b

It doesn't seem to matter which RAW viewer is used. *I viewed the picture
with both Canon's and RawTherapee's. *The posted pic came from the RT capture
to JPEG conversion. *I don't know if the site allows uploads larger than
10Megs.

I'd swear that taking a picture without any lens on it would result in the
same capture. *Explanations?

Well, I blew it up to 500% in Photoshop and I don't see what you are
asking. There is a little noise in the darker areas and less in the
brighter areas -- hopefully to be expected. Other than that it's just an
out of focus photo. Could you restate the question?

OK, good, I'm not the only one. I don't see anything other than a
picture that is very out of focus.

I have no idea what he was trying to show.
--
Tony Cooper - Orlando, Florida

On 03/10/2011 05:54 AM, SneakyP wrote:
wrote in news:a3088e80-dd24-425f-a579-
:

OK, good, I'm not the only one. I don't see anything other than a
picture that is very out of focus.



"CTRL +" the picture (on Firefox browser) until it's obvious the view of a
part of the picture is showing a "mottled" appearance, rather than a smooth
one. It is more visible in the green areas.

A mild gaussian blur would demonstrate the removal of such patterns.

An unsharp mask would reveal the granulation even more.


Is this "mottling" what is called luminance noise?



Oh, and to the poster who spat out an obnoxious diatribe about killfiltering,
blow it out your ass, troll.

Blown up x8 I only see reasonable variation between adjacent pixels. If
your subject has details (foliage, for instance) in the green areas, you
won't get a uniform color even if you defocus.
--
Bertrand

On Thu, 10 Mar 2011 10:26:52 +0100, Ofnuts
wrote:

On 03/10/2011 05:54 AM, SneakyP wrote:
wrote in news:a3088e80-dd24-425f-a579-
:

OK, good, I'm not the only one. I don't see anything other than a
picture that is very out of focus.



"CTRL +" the picture (on Firefox browser) until it's obvious the view of a
part of the picture is showing a "mottled" appearance, rather than a smooth
one. It is more visible in the green areas.

A mild gaussian blur would demonstrate the removal of such patterns.

An unsharp mask would reveal the granulation even more.


Is this "mottling" what is called luminance noise?



Oh, and to the poster who spat out an obnoxious diatribe about killfiltering,
blow it out your ass, troll.

Blown up x8 I only see reasonable variation between adjacent pixels. If
your subject has details (foliage, for instance) in the green areas, you
won't get a uniform color even if you defocus.

LOL!!!! I know exactly what his problem is. But this is much more fun
watching a bunch of silly newsgroup twit trolls try to figure it out.
Photo-editors 101.

LOL!!!!!!!!!!

On Wed, 09 Mar 2011 22:54:13 -0600, SneakyP
wrote:

otter wrote in news:a3088e80-dd24-425f-a579-
:

OK, good, I'm not the only one. I don't see anything other than a
picture that is very out of focus.



"CTRL +" the picture (on Firefox browser) until it's obvious the view of a
part of the picture is showing a "mottled" appearance, rather than a smooth
one. It is more visible in the green areas.


It you are talking about zooming in past a 1:1 pixel display in FF all
you may be doing is showing issues with FF upsampling.

SneakyP wrote:
For those who have some experience with camera picture processing, what is it
that causes differences in the portions of a picture much further than what
would be expected?

The following example is a picture taken of a house across the street, using
a Canon Digital Rebel XS, at ISO 100. The photo was taken out-of-focus on
purpose, to demonstrate the capture of the sensor. It was done to highlight
the graininess seen in pictures when blown up to the pixel level.

What is the reason that this capture shows multiple artifacts? Is it Canon's
particular in-camera RAW capture that does this?

I don't care about the JPG ones as the artefacts also show up in the RAW
file, even before the JPEG conversion. Also never mind the dead-pixel that
shows in the lower right corner of the picture.

http://i172.photobucket.com/albums/w...g?t=1299705139

or just

http://tinyurl.com/4t8b76b

It doesn't seem to matter which RAW viewer is used. I viewed the picture
with both Canon's and RawTherapee's. The posted pic came from the RT capture
to JPEG conversion. I don't know if the site allows uploads larger than
10Megs.

I'd swear that taking a picture without any lens on it would result in the
same capture. Explanations?

It appears you are looking at Photon Noise. Typically
seen in a blue sky area of an image, it is a "natural
phenomena" that cannot be avoided.

Light sometimes acts in ways that can only be described
as a "wave", and at other times in ways that are like
"particles". This is one where it is a particle, called
a photon. Particles called photons land on the sensor
and are counted to determine how much light there was.

But an evenly illuminated area does not get equal
numbers of photons in any given time frame, simply
because photons are discrete particles (if light were
acting like a wave, it would be equal everywhere).

Hence in the time your camera's shutter is open some
sensor locations will count more photons than the
average, and other locations (across an equally
illuminated area) will count less than the average. And
that means that across that area the image will show
variations in brightness from one pixel to another.

In more brightly lit areas there are a lot more photons
being counted, so each pixel is more likely to be very
close to the average but the maximum variation will be
greater. In darker areas where there are fewer photons
to count it is more likely that the actual count will
different from the average by a large relative
percentage, but the maximum variation will be less. The
upshot of that is making dark areas brighter in
photoshop is a good way to see Photon Noise, and so is
making bright areas darker! As you suggested using USM
will make it more obvious, and so will increased
contrast and lowered brightness for the entire image
rather than just around edges.

--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

On 10/03/2011 04:54, SneakyP wrote:
wrote in news:a3088e80-dd24-425f-a579-
:

OK, good, I'm not the only one. I don't see anything other than a
picture that is very out of focus.

"CTRL +" the picture (on Firefox browser) until it's obvious the view of a
part of the picture is showing a "mottled" appearance, rather than a smooth
one. It is more visible in the green areas.

A mild gaussian blur would demonstrate the removal of such patterns.

An unsharp mask would reveal the granulation even more.


Is this "mottling" what is called luminance noise?

Not really. It is from the interaction between thermal noise and sensor
readout noise with the Bayer demosaicing algorithm in the abscence of
any other clues. The result is that a red/green/blue pixel that is hit
by a cosmic ray or is just a bit hot affects neighbouring pixels and on
a very smooth background that can become visible on the right tones.

Scientific imaging CCDs are often actively cooled to minimise the
thermal noise on long exposures but it isn't realistic to do that for a
consumer grade camera. Even then we have to live with cosmic ray hits.

Regards,
Martin Brown

On Thu, 10 Mar 2011 15:16:20 +0000, Martin Brown wrote:

Not really. It is from the interaction between thermal noise and sensor
readout noise with the Bayer demosaicing algorithm in the abscence of
any other clues. The result is that a red/green/blue pixel that is hit
by a cosmic ray or is just a bit hot affects neighbouring pixels and on
a very smooth background that can become visible on the right tones.

The image data is also quantised, (usually to between ten and fourteen
bits) and if the manufacturers are doing their job right this
quantisation will be dithered. Without dithering smooth gradients will
result in visible banding as the pixel level varies between two quantised
values, whereas dithered (noise-added) signals will bump between the two
adjacent quantised values so that the over-all effect is smooth. In
general it is much more important to be smooth at a distance than at a
distance where individual pixels are discernable. Of course, the added
noise that results from the resistive self-noise of the sense amplifiers
may be entirely adequate to provide the dither, rather than having to add
it deliberately in the A/D conversion process.

Cheers,

--
Andrew

Andrew Reilly wrote in news:8tt1qmFpbaU1
@mid.individual.net:

On Thu, 10 Mar 2011 15:16:20 +0000, Martin Brown wrote:

Not really. It is from the interaction between thermal noise and sensor
readout noise with the Bayer demosaicing algorithm in the abscence of
any other clues. The result is that a red/green/blue pixel that is hit
by a cosmic ray or is just a bit hot affects neighbouring pixels and on
a very smooth background that can become visible on the right tones.

The image data is also quantised, (usually to between ten and fourteen
bits) and if the manufacturers are doing their job right this
quantisation will be dithered. Without dithering smooth gradients will
result in visible banding as the pixel level varies between two quantised
values, whereas dithered (noise-added) signals will bump between the two
adjacent quantised values so that the over-all effect is smooth. In
general it is much more important to be smooth at a distance than at a
distance where individual pixels are discernable. Of course, the added
noise that results from the resistive self-noise of the sense amplifiers
may be entirely adequate to provide the dither, rather than having to add
it deliberately in the A/D conversion process.


Thanks for the explanations.


So, in effect, all camera sensors will have this, but in varying degrees
depending upon the brand, model, etc.

I've been wondering if Canon's low end dslr has "added" even more noise to
the picture on purpose. Here is why I ask. People have posted their
pictures from various cameras and from what I've been able to see- the
better the camera handles pixels, the better the end result is. "Larry
Thong" put up some pictures, from a Nikon, that are extremely satisfying to
the eye - but they're converted to low-res or smaller JPEG, but some others
have posted very large pictures and some even RAW ones. It's very
noticeable that those pictures from higher-end cameras don't seem to suffer
the same 'noise' anomalies.

Oh, and if this makes sense- the same thing affected the Canon SD1000,
circa 2005.

I've had other cameras too, but there wasn't the same mottling. i.e.
Fujipix 2Megapixel pictures, for instance.


I'm just wondering if this is the result of a lower-end camera in Canon
lines. That's all.

--
__
SneakyP
To email me, you know what to do.

Supernews, if you get a complaint from a Jamie Baillie, please see:
http://www.canadianisp.ca/jamie_baillie.html