Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
11-09-2012, 05:48 PM   #91
Site Supporter
Site Supporter
Digitalis's Avatar

Join Date: Mar 2009
Location: Melbourne, Victoria
Photos: Gallery
Posts: 11,694
QuoteOriginally posted by vonBaloney Quote
If you print big enough, the sharpness, although we should stop calling it that since that isn't really the difference but the increased detail, clarity, and crispness, yes...
The correct term is acutance.

11-10-2012, 09:24 AM   #92
Site Supporter
Site Supporter




Join Date: Feb 2008
Location: Irvine, CA
Photos: Gallery | Albums
Posts: 307
QuoteOriginally posted by falconeye Quote
As has certainly been said already, this isn't possible. A Bayer-AA filter cannot be simulated in software. Without a Bayer-AA filter, information about the original color of a given light ray is lost forever.

What anti color-moiré tools (in LR and other raw converters) do is very heuristic. Basically, they desaturate (parts of) an image and blur to remove the remaining false moiré in the luminance channel.
Falconeye, I differ slightly with your opinion here, or perhaps I need to state my statement more clearly. A AA-filter can be closely simulated (indeed not perfectly simulated) when applied to the raw light value matrix, (which has no real color information yet, just a pixel luminance data and later the pixel position is assigned a color), prior to the debayering process. So there is minimal processing done to the raw image data to make it possible to closely simulate the AA-filter, that would work well with lower noise images and probably some other better suited conditions.

The anti-moire tools indeed attempt to reduce debayered image moire effects, and is not a AA-filter software simulation by any means, and I completely agree with you here.

In the end, I guess the K5IIs may provide some sharpness benefits with the use of very sharp lenses. I am still trying to figure out practical benefits over the cons. Aggressively cropped photos jumps to mind. Very large prints that are viewed up close would be another perhaps. Astro-photography would practical application for the k5IIs, especially for planetary work... but then $200 webcams do quite well too, so not sure what to think.
11-10-2012, 10:59 AM   #93
Veteran Member
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,653
QuoteOriginally posted by Eric Seavey Quote
Falconeye, I differ slightly with your opinion here, or perhaps I need to state my statement more clearly. A AA-filter can be closely simulated (indeed not perfectly simulated) when applied to the raw light value matrix, (which has no real color information yet, just a pixel luminance data and later the pixel position is assigned a color), prior to the debayering process. So there is minimal processing done to the raw image data to make it possible to closely simulate the AA-filter, that would work well with lower noise images and probably some other better suited conditions.
I'd love to see an example of this magical process which retrieves information which never got recorded.
11-10-2012, 05:53 PM   #94
Veteran Member




Join Date: Jun 2009
Location: Chicago Suburbs.
Photos: Gallery
Posts: 501
perfect solution!

QuoteOriginally posted by Eric Seavey Quote
Falconeye, I differ slightly with your opinion here, or perhaps I need to state my statement more clearly. A AA-filter can be closely simulated (indeed not perfectly simulated) when applied to the raw light value matrix, (which has no real color information yet, just a pixel luminance data and later the pixel position is assigned a color), prior to the debayering process. So there is minimal processing done to the raw image data to make it possible to closely simulate the AA-filter, that would work well with lower noise images and probably some other better suited conditions.

The anti-moire tools indeed attempt to reduce debayered image moire effects, and is not a AA-filter software simulation by any means, and I completely agree with you here.

In the end, I guess the K5IIs may provide some sharpness benefits with the use of very sharp lenses. I am still trying to figure out practical benefits over the cons. Aggressively cropped photos jumps to mind. Very large prints that are viewed up close would be another perhaps. Astro-photography would practical application for the k5IIs, especially for planetary work... but then $200 webcams do quite well too, so not sure what to think.
Just use cheap, crappy lenses with your expensive K-5 IIs. That will be more than enough blur to reduce any moire you get, and you can pocket the savings to boot!



11-10-2012, 05:54 PM   #95
Pentaxian
Class A's Avatar

Join Date: Aug 2008
Location: Wellington, New Zealand
Posts: 11,250
Original Poster
QuoteOriginally posted by Eric Seavey Quote
So there is minimal processing done to the raw image data to make it possible to closely simulate the AA-filter, that would work well with lower noise images and probably some other better suited conditions.
While there is potential to somewhat address moiré at the demosaicing stage already, there is definitively no way to "simulate" an AA-filter. As Falk wrote, the absence of an AA-filter means that some colour information will be lost forever. No subsequent processing can address this.

A moiré aware demosaicing algorithm may account for some scenarios but what if what the algorithm suppresses was actually the real scene? Without an AA-filter, the sensor is recording information it strictly speaking should not be able to record. Imagine a line of small squares with sizes so that they exactly correspond to a sensor sensel. A sensor without an AA-filter will record the same information no matter whether the squares alternate in a black and white pattern or they alternate with respective Bayer filter colours and black. If you arranged the colour squares strategically so that a red square matched a blue sensel, etc., and the colour filters in the Bayer matrix were a 100% discriminating (which they are not) then the AA-filterless sensor would record a black image, despite all the colour patches. No amount of post-processing can fully address this fundamental flaw, i.e., that you don't know what the real information was.

If an AA-filter is used, any spatial information that cannot be truthfully captured by the sensor anyhow is removed so you can fully rely on the information obtained. Due to the Bayer matrix principle, there is still a need for some level of reconstruction guesswork but, a) this is needed in both cases anyhow, and b) will produce, if any, smooth errors, not ones that are obvious at the pixel level (colour artefacts), or even image level (moiré ).

To me, removing an AA-filter is a hack. The better solution is to increase the resolution and keep an AA-filter. This achieves the same resolution advantages without the drawbacks. Of course there are challenges in increasing pixel-pitch, but I'm happy to wait for those to be addressed rather than having "fun" with artefacts from AA-filterless sensors.

Last edited by Class A; 11-10-2012 at 06:03 PM.
11-10-2012, 06:04 PM   #96
Veteran Member




Join Date: Jun 2009
Location: Chicago Suburbs.
Photos: Gallery
Posts: 501
Question about the original topic......

QuoteOriginally posted by Class A Quote
In all these cases it would be better to remove the cause of the problem rather than slightly mitigating its effect by not using an AA-filter.

Note that the (necessary) blurring effect of an AA-filter can be completely undone without loss of information, if the original image has pixel-level sharpness. Studio shots, using flash to stop any motion with good lighting allowing the lens to be used at optimal apertures, for example, can be capture-sharpened to optimal detail levels. Anything a camera without an AA-filter records beyond that, is false detail and can potentially destroy parts of the image.

EDIT: I should have written that my best guess is that "the vast majority is much better served with a complete image forming system." Fast forward to a better argued position by falconeye.
First, a great post from Class A. Thanks for your many contributions to PF.

You say the blurring effect can be reversed with capture-sharpening. Can you post pictures that demonstrate what you are talking about? What would the theoretical amount of sharpening be with common forms of moire (AA mitigated)? Would one use USM or High Pass Filter or some other magic elixer??? I mean, after all, others have been posting how fantastic the lack of an "AA" type filter is with pictures posted. Can you do the same??? Supposedly, the moire can be predicted, then the opposite must be true also, that the effects of the AA filter can also be predicted and a formula for the type of capture sharpening needed to get rid of the AA effects. I am just trying to see if there is a way to prove the point you are making in the original post.


Last edited by goldenarrow; 11-10-2012 at 06:09 PM. Reason: clarification
11-11-2012, 06:43 AM   #97
Site Supporter
Site Supporter




Join Date: Feb 2008
Location: Irvine, CA
Photos: Gallery | Albums
Posts: 307
QuoteOriginally posted by falconeye Quote
As has certainly been said already, this isn't possible. A Bayer-AA filter cannot be simulated in software. Without a Bayer-AA filter, information about the original color of a given light ray is lost forever.

What anti color-moiré tools (in LR and other raw converters) do is very heuristic. Basically, they desaturate (parts of) an image and blur to remove the remaining false moiré in the luminance channel.


QuoteOriginally posted by Class A Quote
While there is potential to somewhat address moiré at the demosaicing stage already, there is definitively no way to "simulate" an AA-filter. As Falk wrote, the absence of an AA-filter means that some colour information will be lost forever. No subsequent processing can address this.

A moiré aware demosaicing algorithm may account for some scenarios but what if what the algorithm suppresses was actually the real scene? Without an AA-filter, the sensor is recording information it strictly speaking should not be able to record. Imagine a line of small squares with sizes so that they exactly correspond to a sensor sensel. A sensor without an AA-filter will record the same information no matter whether the squares alternate in a black and white pattern or they alternate with respective Bayer filter colours and black. If you arranged the colour squares strategically so that a red square matched a blue sensel, etc., and the colour filters in the Bayer matrix were a 100% discriminating (which they are not) then the AA-filterless sensor would record a black image, despite all the colour patches. No amount of post-processing can fully address this fundamental flaw, i.e., that you don't know what the real information was.

If an AA-filter is used, any spatial information that cannot be truthfully captured by the sensor anyhow is removed so you can fully rely on the information obtained. Due to the Bayer matrix principle, there is still a need for some level of reconstruction guesswork but, a) this is needed in both cases anyhow, and b) will produce, if any, smooth errors, not ones that are obvious at the pixel level (colour artefacts), or even image level (moiré ).
In all fairness, I thought about this some more, and will take my comments back regarding simulating the AA filter at the raw level. Yes, my description would work fine with white light, but mess up colored bands of light, where such simulation would possibly discard color and cause other problems. So the light path order (AA-filter) needs to be: AA-blurring, color filter (=bayer), then demozaicing. Indeed slightly blurring the image in software after passing through the color filters is not the same at all.

Sorry sloppy thinking on my part, and I will admit my thoughts were not accurate. Though I studied physics, specialized in acoustics, I like facts and I will have to admit my bit embarrassing error that became so clear after looking into this more thoroughly....sorry folks.


Last edited by Eric Seavey; 11-11-2012 at 07:20 AM.
11-11-2012, 08:27 AM   #98
Veteran Member
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,653
QuoteOriginally posted by falconeye Quote
As has certainly been said already, this isn't possible. A Bayer-AA filter cannot be simulated in software. Without a Bayer-AA filter, information about the original color of a given light ray is lost forever.
One thing that I am puzzled by is the 'weak' AA filter. What advantage are you gaining by not matching the birefringent image offset to the sensel spacing, if that is indeed what is happening ?
11-11-2012, 04:14 PM   #99
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
QuoteOriginally posted by kh1234567890 Quote
One thing that I am puzzled by is the 'weak' AA filter. What advantage are you gaining by not matching the birefringent image offset to the sensel spacing, if that is indeed what is happening ?
AFAIK, there is no good article looking into this.

My understanding is based on two observations and then I extrapolate:
  1. An offset of x% of the sensel spacing (pixel pitch) is easily manufactured (just make the filter x% as thin). And at x=0, we have the case of no Bayer-AA filter, at x=100% we have a full Bayer-AA filter. So, any intermediate case is easily doable. The result could be a weighted overlay of both both images but in practice, it is likely more complex. And note that some blurring does always occur.
  2. Even with a full Bayer-AA filter, the colors of the demosaiced image aren't well defined. The problem isn't the resolution of the full image. The problem is that even 2x2 blurred rays don't yield a clear half-resolution RGB image. Imagine you'd combine each 2x2 Bayer cell into one RGB pixel. You then would end up with images which have very ugly fringes along edges. Examples exist in the net from people who did the exercise. Therefore, some heuristic image reconstruction must be applied to construct an RGB image with possibly no false colors.
And this heuristic algorithm is supported by blurred color information (this is an assertion to the algorithm that color information cannot change quicker than with a given maximum spatial frequency). This assertion makes the result more reliable. But any degree of possible assertion should help the algorithm.

BTW: Current demosaicing algorithms aren't that smart. They are heuristic rather than model-based. I assume that a smarter demosaicing algorithm could be written which avoids false color moiré patterns if assertions cannot support it. The demosaicer then would have to know (or detect) the degree of blur in an image (and apply an auto-adjusted amount of capture sharpening automatically).
11-14-2012, 06:14 AM   #100
Veteran Member




Join Date: Jun 2009
Location: Chicago Suburbs.
Photos: Gallery
Posts: 501
Let's try again - how to get rid of AA blur?

QuoteOriginally posted by goldenarrow Quote
First, a great post from Class A. Thanks for your many contributions to PF.

You say the blurring effect can be reversed with capture-sharpening. Can you post pictures that demonstrate what you are talking about? What would the theoretical amount of sharpening be with common forms of moire (AA mitigated)? Would one use USM or High Pass Filter or some other magic elixer??? I mean, after all, others have been posting how fantastic the lack of an "AA" type filter is with pictures posted. Can you do the same??? Supposedly, the moire can be predicted, then the opposite must be true also, that the effects of the AA filter can also be predicted and a formula for the type of capture sharpening needed to get rid of the AA effects. I am just trying to see if there is a way to prove the point you are making in the original post.

I would like to continue the conversation by trying to show that there might be a way to use capture sharpening software to mitigate the effects of the AA filter. But has anyone measured what kind of sharpening is needed, as in specific radius, strength, threshold settings (or other software tricks) for a particular AA filter? Or is that a professional secret??? Perhaps it needs someone who has access to both a K-5 II and a K-5 IIs using the SAME lens on a target and then publish the results. I think it would be a great study and prove that one does not have to buy the IIs model to achieve just-as-sharp results. Just saying this because most of this thread seems to be theoretical and I would love to see a more demonstrative proof.

11-14-2012, 09:41 PM   #101
Veteran Member




Join Date: Jun 2009
Location: Chicago Suburbs.
Photos: Gallery
Posts: 501
dpreview photo comparison

QuoteOriginally posted by goldenarrow Quote
I would like to continue the conversation by trying to show that there might be a way to use capture sharpening software to mitigate the effects of the AA filter. But has anyone measured what kind of sharpening is needed, as in specific radius, strength, threshold settings (or other software tricks) for a particular AA filter? Or is that a professional secret??? Perhaps it needs someone who has access to both a K-5 II and a K-5 IIs using the SAME lens on a target and then publish the results. I think it would be a great study and prove that one does not have to buy the IIs model to achieve just-as-sharp results. Just saying this because most of this thread seems to be theoretical and I would love to see a more demonstrative proof.

Just looking at the DP review photos of the High ISO Pentax K-5 II compared to the K-5 IIs and the superiority of the K-5 IIs is noticeable at all ISOs. Makes it hard to think of a way to simulate the lack of AA filter with a camera that has an AA filter, like the K-5 II. The detail is too delicate for sharpening to work in a subtle way to replicate, unless one knew ahead of time what types of sharpening are needed to get it to look the same way for that model camera's AA filter. The IIs camera seems to have the edge in all the comparisons. Are there any sharpened photo comparisons with an AA-filtered camera that look as close as possible to the K-5 IIs ??? And how much work would it take (hours and hours???)?

If there is no answer, then I will have to assume that the answer is "no", there is no comparison done (yet) to get the look of the K-5 IIs. And maybe there never will be. Proof is in the pudding, as they used to say........

11-14-2012, 09:45 PM   #102
Pentaxian




Join Date: Oct 2011
Location: Albuquerque, NM
Posts: 6,027
QuoteOriginally posted by goldenarrow Quote
Just looking at the DP review photos of the High ISO Pentax K-5 II compared to the K-5 IIs and the superiority of the K-5 IIs is noticeable at all ISOs. Makes it hard to think of a way to simulate the lack of AA filter with a camera that has an AA filter, like the K-5 II. The detail is too delicate for sharpening to work in a subtle way to replicate, unless one knew ahead of time what types of sharpening are needed to get it to look the same way for that model camera's AA filter. The IIs camera seems to have the edge in all the comparisons. Are there any sharpened photo comparisons with an AA-filtered camera that look as close as possible to the K-5 IIs ??? And how much work would it take (hours and hours???)?

If there is no answer, then I will have to assume that the answer is "no", there is no comparison done (yet) to get the look of the K-5 IIs. And maybe there never will be. Proof is in the pudding, as they used to say........

You can't create detail that was never captured out of thin air.
11-14-2012, 10:14 PM - 1 Like   #103
Veteran Member
JohnBee's Avatar

Join Date: Sep 2008
Location: Newrfoundland
Photos: Albums
Posts: 4,667
QuoteOriginally posted by goldenarrow Quote
...Are there any sharpened photo comparisons with an AA-filtered camera that look as close as possible to the K-5 IIs ??? And how much work would it take (hours and hours???)?
Pentax K-5 IIs vs Pentax K-5 II (with deblurr).
Total time to process: It took me a bit to figure out what worked best, though applying it now, is as simple as clicking OPEN in RAW developer.

The trick is to apply a deblurr adjustment to the K-5 II image. OR... in this particular case, what is called Deconvoluted Sharpening in Raw Therapee. Once this is done, the K-5 II is pretty much on equal footing with the K-5 II in terms of detail though it does benefit from the lack of moire and aliasing in the end.

PS. there is room for sharpening improvements on the K-5 II, though I think this pretty much sums it up.

Last edited by JohnBee; 11-14-2012 at 10:46 PM.
11-14-2012, 10:28 PM   #104
Veteran Member
JohnBee's Avatar

Join Date: Sep 2008
Location: Newrfoundland
Photos: Albums
Posts: 4,667
Yes they are labeled correctly
11-14-2012, 10:29 PM   #105
Pentaxian




Join Date: Oct 2011
Location: Albuquerque, NM
Posts: 6,027
QuoteOriginally posted by JohnBee Quote
Yes they are labeled correctly
Oh, you caught me (I deleted my message). User error -- had my tabs mixed up -- moire is in the IIs version...
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
aa-filter, blur, camera, cause, colour, detail, dslr, image, information, k-5, photography, sharpness
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
From K20D to 645D, Pros and Cons. Reportage Pentax DSLR Discussion 12 11-10-2010 03:13 PM
Pentax K-5 pros/cons psychdoc Pentax K-5 & K-5 II 11 11-08-2010 12:41 PM
Decision: Pentax K10D, K20D, or K7? Pros and cons? Lulerfly Pentax DSLR Discussion 32 11-04-2010 09:14 AM
Pros and cons of the Pentax KM/2000 lesmore49 Pentax DSLR Discussion 109 05-01-2009 11:51 PM
Sigma 28-70 F/2.8 (Pros & Cons) GLThorne Pentax SLR Lens Discussion 8 03-24-2007 07:10 AM



All times are GMT -7. The time now is 07:54 AM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top