Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
12-08-2015, 07:48 AM   #31
Pentaxian




Join Date: Jan 2011
Location: New York
Posts: 3,276
I think some people above might be confused between an anti-aliasing and Bayer filter.

A Bayer filter is needed to get color with the Sony sensors in Pentax cameras. The sensor is monochrome and only measures brightness, and the Bayer filter is a color overlay with a specific green/red/blue pattern. (Some non-Sony sensors like Fuji X-Trans also use a color filter even though it's not a "Bayer" pattern; same concept different design)

The Bayer filter and subsequent calculations to get a color image can sometimes trigger false colors in fine patterns (clothing thread, distant fences, etc.).

An anti-aliasing (AA) filter adds a slight blur to lessen the false color from a Bayer filter. As pixel counts increase, there's less risk of the false color, and the AA filter becomes less important.

A "strong" AA filter creates more blur than a "weak" AA filter. The strong filter has less risk of false color, at the expense of a less sharp image.

12-08-2015, 08:34 AM   #32
Veteran Member




Join Date: Aug 2010
Location: Houston, Texas
Posts: 757
QuoteOriginally posted by Nicolas06 Quote
As for sharpening, this add artefacts if abused and can only applied for a limited effect... Sharpening doesn't allow a 55-300 to have the level of detail of DA300. This cannot repeated and tend to kill the bokeh too. In fact one of the things I really like is that I don't need to sharpen results now anymore and I don't get the associated artefacts. This give me much better details that look more natural.

Thanks for that. As I sharpen I sometimes notice that I introduce artifacts that I'd rather not have. I probably need to take a sharpen tutorial, as there is probably a lot to be considered in this step.
12-08-2015, 08:46 AM   #33
Pentaxian
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,405
QuoteOriginally posted by DeadJohn Quote
A "strong" AA filter creates more blur than a "weak" AA filter. The strong filter has less risk of false color, at the expense of a less sharp image.
That is rubbish. The AA filter creates four slightly offset images, the offset distance being essentially set by the thickness of the birefringent material (and its refractive index). How do you make this 'weak' ?
12-08-2015, 12:25 PM   #34
Pentaxian




Join Date: Nov 2013
Posts: 4,686
QuoteOriginally posted by kh1234567890 Quote
That is rubbish. The AA filter creates four slightly offset images, the offset distance being essentially set by the thickness of the birefringent material (and its refractive index). How do you make this 'weak' ?
You have written the response already.

12-08-2015, 12:35 PM   #35
Pentaxian




Join Date: Nov 2013
Posts: 4,686
QuoteOriginally posted by GlennG Quote
Thanks for that. As I sharpen I sometimes notice that I introduce artifacts that I'd rather not have. I probably need to take a sharpen tutorial, as there is probably a lot to be considered in this step.
I found that if the original image is sharp, you don't need at all to add a sharpness filter. I also found that a sharpness filter by default tend to concentrate on a given level of detail (often very small detail at pixel level) rather than increasing overall constrast. You can of couse tweak it, but the tweak tend to improve one spacial frequency range only.

My pictures tend to have lot of definition out of the box because I tend to use sharp lenses and a filterless sensor (here a K3). The post processing software DxO also tend to use the best compromize for its raw conversion based on lens profile. There no point to add more sharpness, because really, it add no value, at least in my case.

What I work with is contrast and micro constrast depending of the kind of image I want. This allow me to get what I want without any artifact except for some extreme value of micro contrast.
12-08-2015, 12:58 PM - 1 Like   #36
Site Supporter




Join Date: Mar 2009
Location: Gladys, Virginia
Photos: Gallery
Posts: 16,017
John Bee compared the K5 II and K5 IIs. You can argue about his techniques back then, but the end result, to my eyes was that after appropriate sharpening techniques the images were extremely close -- with the K5 II having less moire.

https://www.pentaxforums.com/forums/115-pentax-k-5/205185-k-5-iis-vs-k-5-ii-c...ifference.html
12-08-2015, 12:59 PM   #37
Pentaxian




Join Date: Jan 2011
Location: New York
Posts: 3,276
QuoteOriginally posted by Nicolas06 Quote
SNIP...As for sharpening, this add artefacts if abused and can only applied for a limited effect... Sharpening doesn't allow a 55-300 to have the level of detail of DA300. This cannot repeated and tend to kill the bokeh too. In fact one of the things I really like is that I don't need to sharpen results now anymore and I don't get the associated artefacts. This give me much better details that look more natural.
Lightroom's "masking" slider on the detail panel helps protect bokeh. Hold the alt key while moving the slider to see which portions of the image will not be sharpened. Other sharpening software likely has a similar setting, without resorting to the extra work of layer masks.
12-08-2015, 01:06 PM   #38
Pentaxian
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,405
QuoteOriginally posted by Nicolas06 Quote
Originally posted by kh1234567890
That is rubbish. The AA filter creates four slightly offset images, the offset distance being essentially set by the thickness of the birefringent material (and its refractive index). How do you make this 'weak' ?
You have written the response already.
What possible advantage would there be by making the offset distance different from the sensel spacing ?

12-08-2015, 01:58 PM   #39
Pentaxian




Join Date: Nov 2013
Posts: 4,686
QuoteOriginally posted by kh1234567890 Quote
What possible advantage would there be by making the offset distance different from the sensel spacing ?
If you have a fully effective low pass filter and colors are fully merged, you basically get 1/4 of the resolution. You actually quite blur the picture. So a so called "16MP" is only really 4MP with full color information and because all the light was spreat the luminance information is also quite blured.

If you use a weak filter with less than the distance between 2 photosites, this mean that a given photosite still get more light from it's own spacial position than from the light of its neigthbourgh. This increase the resolution. In particular:
- for white/grey that pass through all color filters
- for green that represent half of the pixels and not only 1/4.
- for all colors because colors filter don't fully filter other colors but only do it partially
12-08-2015, 02:15 PM   #40
Pentaxian




Join Date: Nov 2013
Posts: 4,686
I would add that this whole low pass filter is necessary based on naive application of the theory. It is indeed true that with a 16M photosite sensor with only 1 color per photosite arranged in square of 4 photosites you have only 4MP with full color information and so that you could, in theory represent signal accurately not greater than half this spacial frequency so no more than 1MP.

This would be the case for a purely random signal. A signal without any predictible shape and no usuful information.

But this is not how its work. Interresting signals have given characteristics. They have predicitive gradiant, pattern and shape. You can extrapolate/guess the information from experience and nearby photosites. This is exactly how compression work and that's why a JPEG with a compression factor of something like 20 is almost not distinguishable from the original picture. This is just we are not interrested in the exact signal or to record every possible signal, but only signal that make sense and that are likely to actually happen in reality.

That what a raw converter is doing using all the information as its disposal:
- the actual pattern of colors filters and photosites.
- the color sensitivity curves of each color filter... The filter are not binary. they let pass some light from other colors, only they are less sentivite to them.
- knowledge of how a typical meaningfull signal look like.
- lot of fine tuning with empirical trial and errors.

This is what allow to get "almost" 16MP worth of detail and not just 1MP pr even 4MP for the worst case senario with an heavy low pass filter. I say almost because pixel shift still look significantly better, showing that there margin for progress. But we get much better result than a naive application of the theory would allow.

Using a weaker low pass filter (or no low pass filter at all) allows the algorithm to better predict the signal that has likely genered this recording at the expense of being more worng on the occasion the guess is wrong. That's a tradeoff.

And there no better/worse tradoff, there just tradeoffs and actual experience dealling with them with settings/configuration that tend to work better in practice for actual pictures. Not for every possible signal that we could wish to record in theory but we will never encounter.

People only complained when the announcement was made offical that we removed low pass filters but this is forgetting that the years before the manufacturers experimented with different low pass filter design, weaker and weaker as they have seen that overall it performed better. All of this didn't overnigh with engineer using full low pass filter for 10 years then changing their mind and no using low pass filter at all anymore.

Class A think he know better than the engineers. He think that obviouly if they do it, it is not because it is a better tradeoff but only to please consumers. That may very well be true as the key factor is obviously to sell camera not just to record the best picture in theory. But there no proof there, this is just supposition on both sides.

The best maybe is for everybody to judge from himself, to look at what reviewers the community overall says and what reviewers says. To me I see a benefit and I have taken dozen thousand pictures with camera with and without low pass filter. DxO see a benefit for lens sharpness. Constructors tend to promote it for their high end DSLR.

But sure everybody is legitimate to draw its own conclusions.

Last edited by Nicolas06; 12-08-2015 at 02:34 PM.
12-08-2015, 02:46 PM   #41
Pentaxian
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,405
QuoteOriginally posted by Nicolas06 Quote
If you use a weak filter with less than the distance between 2 photosites, this mean that a given photosite still get more light from it's own spacial position than from the light of its neigthbourgh. This increase the resolution. In particular:
- for white/grey that pass through all color filters
- for green that represent half of the pixels and not only 1/4.
- for all colors because colors filter don't fully filter other colors but only do it partially
Sorry, but you've just set off my bullsh*t alarm ...

References ?
12-08-2015, 02:56 PM   #42
Pentaxian




Join Date: Nov 2013
Posts: 4,686
An exemple of picture without low pass filter:

- In attachment:
- the full picture resized to websize (from 24MP K3 and 50-135 @80mm f/5.6, 1/250s)
- A 100% crop. This level of crop give an equivalent framing of 470mm on APSC or 707mm on FF. It still look okish at web size!

Last edited by Nicolas06; 01-31-2017 at 02:03 PM.
12-08-2015, 03:05 PM   #43
Pentaxian




Join Date: Nov 2013
Posts: 4,686
QuoteOriginally posted by kh1234567890 Quote
Sorry, but you've just set off my bullsh*t alarm ...

References ?
You could read the wikipedia article on demosaicing: https://en.wikipedia.org/wiki/Demosaicing. It far from being a naive interpolation.

Also if DxO use profile for camera to calibrates colors that not just for fun. They adapt to the whole sensor stack used. I think if you are really interrested you could read the source code of the algorithms used in raw therapee. The program is open source.

A guy show an example on how a naive algorithm would perform in practice: http://photo.stackexchange.com/questions/22596/what-are-the-pros-and-cons-of...ing-algorithms

We can see much more aliasing, banding and false colors that much worse than a typical jpeg look like out of camera:

Last edited by Nicolas06; 12-09-2015 at 01:27 AM.
12-08-2015, 04:42 PM   #44
Pentaxian




Join Date: Jan 2011
Location: New York
Posts: 3,276
QuoteOriginally posted by kh1234567890 Quote
Sorry, but you've just set off my bullsh*t alarm ... References ?
Call my post "rubbish", then "bullsh*t" on another poster. Maybe I'm cranky today but here's a word I haven't used in ages: Plonk.
12-09-2015, 01:25 AM   #45
Pentaxian
kh1234567890's Avatar

Join Date: Sep 2010
Location: Manchester, UK
Photos: Gallery
Posts: 2,405
QuoteOriginally posted by DeadJohn Quote
Plonk
Oh well, so you won't be interested in knowing that it was actually an improvement in the sensor microlenses that enabled the manufacturers to make more profit by leaving out the birefringent stack AA filter ...
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!
camera, dslr, photography, sensor
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Is ther a K-x or other AA Pentax for sale on the Forum kamicota Pentax DSLR Discussion 21 08-17-2015 07:33 AM
Is a K-3 with a full frame sensor enough for you? jake14mw Pentax Full Frame 58 03-28-2015 11:39 AM
Pentax staying small. An advantage? stillshot2 Photographic Industry and Professionals 27 12-22-2013 02:34 AM
Bayer filtered camera sensor brian101 Troubleshooting and Beginner Help 4 05-08-2013 01:10 PM
What is the advantage of 21/35/40 prime vs a f2.8 zoom lens raider Troubleshooting and Beginner Help 23 06-21-2010 10:57 AM



All times are GMT -7. The time now is 01:03 PM. | 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