Sign in with
Sign up | Sign in
Your question

The two pixel shift filter

Last response: in Digital Camera
Share
Anonymous
May 26, 2005 3:56:30 PM

Archived from groups: rec.photo.digital.slr-systems (More info?)

Digital cameras use a phase shift filter in front
of the CCD/CMOS sensor that apparently spreads the
image across two pixels, to eliminate moire and
aliazing(sp?). Is there any software out there that
can reverse it? Reason being I'd like to see if
the improved sharpness on some images would improve
them. I understand they do it to
eliminate problems with images that contain straight
lines but it might not be needed for other kinds of
images.
-Rich

More about : pixel shift filter

Anonymous
May 26, 2005 3:56:31 PM

Archived from groups: rec.photo.digital.slr-systems (More info?)

RichA wrote:

> Digital cameras use a phase shift filter in front
> of the CCD/CMOS sensor that apparently spreads the
> image across two pixels, to eliminate moire and
> aliazing(sp?). Is there any software out there that
> can reverse it? Reason being I'd like to see if
> the improved sharpness on some images would improve
> them. I understand they do it to
> eliminate problems with images that contain straight
> lines but it might not be needed for other kinds of
> images.

Does that actually show up in RAW files? DCRAW is an alternate converter
that softens the bayer interpolation a lot less but maybe that doesn't
touch the antialiasing factor.


--
Paul Furman
http://www.edgehill.net/1
san francisco native plants
Anonymous
May 27, 2005 1:00:35 AM

Archived from groups: rec.photo.digital.slr-systems (More info?)

"Paul Furman" <paul-@-edgehill.net> wrote in message
news:A_idnc3cXJeKawjfRVn-rQ@speakeasy.net...
> RichA wrote:
>
>> Digital cameras use a phase shift filter in front
>> of the CCD/CMOS sensor that apparently spreads the
>> image across two pixels, to eliminate moire and aliazing(sp?). Is
>> there any software out there that
>> can reverse it? Reason being I'd like to see if
>> the improved sharpness on some images would improve
>> them. I understand they do it to eliminate problems with images
>> that contain straight
>> lines but it might not be needed for other kinds of
>> images.
>
> Does that actually show up in RAW files? DCRAW is an alternate
> converter that softens the bayer interpolation a lot less but maybe
> that doesn't touch the antialiasing factor.

Different Raw converters may employ different strategies in weighting
the various trade-offs; more-or-less color aliasing or sharpness or
noise reduction or dynamic range or ... etc.

Most converters produce similar results with regards to sharpness, and
that is that they do very little sharpening in their lowest sharpness
setting (exceptions are Pixmantec's Rawshooter 2005, and Capture One,
they apply more sharpening than most). One benefit is that these
images create fewer artifacts when you resize (in particular
downsample) them. Another benefit is that it allows the (skilled) user
to sharpen to his/her hearts content, either for screen display (which
tolerates less sharpening) or for printing (which tolerates more
sharpening depending on printer parameters). Sharpening of already
sharpened images creates more visible artifacts very easily.

It is also possible to strike a compromise, by applying some "capture
sharpening" to compensate for the Anti-Aliasing filter and lens
losses, and by later (after resampling for a particular use) applying
creative and output sharpening. Photoshop CS2 has added a "Smart
Sharpening" functionality to, amongst others, tackle lens blur and its
effect comes close to what more involved techniques produce. But
again, it will make resizing more artifact prone.

Bart
Related resources
Anonymous
May 27, 2005 10:49:27 AM

Archived from groups: rec.photo.digital.slr-systems (More info?)

Bart van der Wolf wrote:
>
>
> It is also possible to strike a compromise, by applying some "capture
> sharpening" to compensate for the Anti-Aliasing filter and lens losses,
> and by later (after resampling for a particular use) applying creative
> and output sharpening. Photoshop CS2 has added a "Smart Sharpening"
> functionality to, amongst others, tackle lens blur and its effect comes
> close to what more involved techniques produce. But again, it will make
> resizing more artifact prone.
>
> Bart

According to Canon's information regarding 10D images applying an
unsharp mask at 300% with 0.2 radius will offset the effects of their
anti-alias filter. 20Ds have less aggressive anti-alias (to their
detriment IMO) and may achieve the same results at 200%

Interestingly enough, images from 20Ds contain more "damaged data" than
the 10D images did. Despite high level approaches to Canon over this
oddity, there is no information forthcoming from them on the subject.

My own opinion is that the increased pixel density of the newer sensor
splats the image over a larger sample of pixels than the earlier on did.
I know I can get big enlargements clearer, from 10D files shot in winter
than from 20D ones taken in high summer.

There must be significant image degradation as the temperature rises
towards 40c - the maximum environment temperature for 20Ds. A themometer
inside the battery case said it was 43c when the outside temperature was
39c. I would think attention to keeping a (Canon) camera cool will
produce more gain than killing the anti-alias filter.

Douglas
Anonymous
May 27, 2005 10:49:28 AM

Archived from groups: rec.photo.digital.slr-systems (More info?)

"Ryadia@Home" <ryadia@hotmail.com> wrote:

>Interestingly enough, images from 20Ds contain more "damaged data" than
>the 10D images did. Despite high level approaches to Canon over this
>oddity, there is no information forthcoming from them on the subject.


And the EOS 350D is even worse than the 20D in this respect.
Anonymous
May 27, 2005 10:49:28 AM

Archived from groups: rec.photo.digital.slr-systems (More info?)

"Ryadia@Home" <ryadia@hotmail.com> wrote in message
news:4296365a$1@dnews.tpgi.com.au...
SNIP
> According to Canon's information regarding 10D images applying an
> unsharp mask at 300% with 0.2 radius will offset the effects of
> their anti-alias filter. 20Ds have less aggressive anti-alias (to
> their detriment IMO) and may achieve the same results at 200%

The problem with generalizations, like this one, is that it is a
generalization and that doesn't take the combined "lens + AA-filter +
sensor sampling density" MTF into consideration. The amount:300%,
Radius:0.3 (!) sharpening suggestion I saw for the 1D series doesn't
look too bad (even though there is room for improvement with more
tuned approaches, based on Point Spread Functions and MTFs).

> Interestingly enough, images from 20Ds contain more "damaged data"
> than the 10D images did.

In relation to the topic at hand, I don't see that confirmed in my MTF
measurements (unless one relies on in camera JPEG conversions alone,
or disregards the larger number of pixels which requires slightly less
magnification). I do see differences between various Raw converter
results (DPP with a Sharpening set to 4 or 5 looks technically very
good (assuming no resizing is needed) compared to the alternatives).

> Despite high level approaches to Canon over this oddity, there is no
> information forthcoming from them on the subject.

???

> My own opinion is that the increased pixel density of the newer
> sensor splats the image over a larger sample of pixels than the
> earlier on did.

Correct, but that is an advantage because it gives more control over
the somewhat oversampled (relative to the 10D) image.

> I know I can get big enlargements clearer, from 10D files shot in
> winter than from 20D ones taken in high summer.

Something doesn't sound right, unless the thermal difference causes
issues (which it does for noise).

> There must be significant image degradation as the temperature rises
> towards 40c - the maximum environment temperature for 20Ds. A
> themometer inside the battery case said it was 43c when the outside
> temperature was 39c. I would think attention to keeping a (Canon)
> camera cool will produce more gain than killing the anti-alias
> filter.

The temperature factor causing more noise (general physics at work),
is quite a different issue than low-pass filtering (human intervention
at work). I wouldn't know how to rate the difference other than in a
rather subjective way ...

Bart
Anonymous
May 27, 2005 11:46:27 PM

Archived from groups: rec.photo.digital.slr-systems (More info?)

"Bart van der Wolf" <bvdwolf@no.spam> wrote in message
news:42964f41$0$155$e4fe514c@news.xs4all.nl...
>
> "Ryadia@Home" <ryadia@hotmail.com> wrote in message
> news:4296365a$1@dnews.tpgi.com.au...
> SNIP
>> According to Canon's information regarding 10D images applying an unsharp
>> mask at 300% with 0.2 radius will offset the effects of their anti-alias
>> filter. 20Ds have less aggressive anti-alias (to their detriment IMO) and
>> may achieve the same results at 200%
>
> The problem with generalizations, like this one, is that it is a
> generalization and that doesn't take the combined "lens + AA-filter +
> sensor sampling density" MTF into consideration. The amount:300%,
> Radius:0.3 (!) sharpening suggestion I saw for the 1D series doesn't look
> too bad (even though there is room for improvement with more tuned
> approaches, based on Point Spread Functions and MTFs).
>
>> Interestingly enough, images from 20Ds contain more "damaged data" than
>> the 10D images did.
>
> In relation to the topic at hand, I don't see that confirmed in my MTF
> measurements (unless one relies on in camera JPEG conversions alone, or
> disregards the larger number of pixels which requires slightly less
> magnification). I do see differences between various Raw converter results
> (DPP with a Sharpening set to 4 or 5 looks technically very good (assuming
> no resizing is needed) compared to the alternatives).

.......
> at work). I wouldn't know how to rate the difference other than in a
> rather subjective way ...
>
> Bart

Apparently DPP uses edge sharpening (I think this is where you detect the
edges with a high pass filter, blur then use this to mask the unsharp mask
operation), this might be the same thing as smart sharpening.
!