No matter the packaging, it's still a dog turd ... ; - )

Cool information about audio sampling - I appreciate you passing on the knowledge, as I love to learn.
Yes, but I suggest to acquire knowledge better from sb already mastered the basics, please not from me. Oversampling in audio has another reason and without knowing the facts (cost and effort reasons, marketed as a superior improvement to the crowd) should not be used as an argument here...

Whether an antiliasing filter is needed or not, everybody can decide by himself on hard facts, much more accurate than anybody else can do for him: Look at the picture, if there are no disturbing artefacts an antialiasing filter was not needed, if there are, the lack of an antialiasing filter has impaired the picture.

The observation of many people here is, that the AA filter has an impact on the majority of the Bayer pictures, and they dislike it. Milling down the AA filter on Bayer cameras gives better results for some enthusiasts. In the professional field the Hasselblads etc. are delivered w/o AA filter, if required it can be used as an option. The Leica Digital Modul R, the Leica M series as well as the Leica S2 does not utilize an AA filter.

One of the main reasons to use AA filters with Bayer cameras is, that the red and blue channel does degrade already at lower frequencies compared to the green channel, there are 2x more green pixels available than red or blue. This can become visible as very distractive color aliasing. The Foveon sensor has all color information for every pixel and therefore does not show this effect. Even if aliasing occurs, there is no color aliasing on Foveon pictures. Color aliasing is ways more disturbing.

For the same picture and lens, the aliasing artefacts will be higher for a sensor with less pixels. Contrary to random noise characterised by the same power for all frequencies real life pictures are bandlimited and carry less power for the higher frequencies. For our eye the residual errors are oven more than acceptable.

Concerning this, our ear is much more sensitive compared to our eye. Our ear can separate higher and lower frequencies very well. Aliasing in the audio field appears as additional single tones or "digital noise" which tends to make instruments sound harsh. In the analoge age another necessisity for an audio amplifier was discovered - the linearity. If a signal is amplified by the same factor for all levels no distortion occures. Nonlinearities produce additional frequency components at multiples of the base frequency. The difference between the tube amps and transistor amps is that distortions for the tube are at hamonics (even freq), contrary the transistor's distortions adds uneven and therefore alien frequency components to real instruments. (For the same reason tube amps are used for electric guitars with the attention to add harmonic distortion to the signal). Another unwanted effect is hard clipping of the signal.

So, why this journey into the audio field? Damnd, who cares about linearity in digital picture processing? Now as we know that non-linearity adds new frequencies to the real signal, we should not want this. The truth is, it does not become visible to our eye, and the first what we do is no manipulate curves by adjusting the light temperature, the contrast and brightness, and do other manipulations for each color, then sharpen the image, adapt the color room!!! For nonlinear editing we use a lot of money and time while in the audio world a lot of money is spent for equipment which only transports the signal, and does not add or take off something from it.

(continued)
 
...continued

To summarize, what we normally do with picture processing is far from processing the reality - as long we don't target perfect reproductions. Everything is allowed, and the photographers are proud of interpreting the picture by their nonlinear processing skills. That a picture might have some unnormality based on nonlinear processing with the intention to improve the image or antialiasing which does not become visible - who cares, really?

Roland-style of argumentation is: "With your Porsche you can't transport a refrigerator!" People get tricked by this idea, they experienced driving a Porsche is so joyful but suddenly they feel they are missing something by not being able to transport a refrigerator! A situation which arises once every 5 years! Even a cheap Lada Niva http://de.wikipedia.org/wiki/Lada_Niva is, according to Roland's logic, superior to a Porsche! Once he got a ride in a Porsche and he realised, in fact, it was more joyful than driving a Lada Niva, how can that be? As a Porsche cannot transport a refrigerator the driving experience should be very bad. His theory needed some correction and he figured out, driving a Porsche intensifies normal joyness beyond reality and should therefore not be there!

"But this is not the point", does Sandy argue again and again and again, "use the right tool, just rent a van if you need a van and use your main tool most of the time, and enjoy".

Man, some people can be so stur, not moving one micron from their point of view. More and more people here admit that they would prefer a more responsive camera with better auto focus and image stabilisation and video and battery life and Canon mount, the list of wishes is even longer. They even say they would not mind or even prefer to buy a Foveon based NiCan, even they say as long it is a full color sensor they don't mind Foveon, either. The only reason they have invested in Sigma cameras are the image quality which miss the digital look of other cameras. This is a personal preference, nothing absolute. Roland and many other don't see the difference, but this is not a proof that it is not there...

Even worse, with Roland's argumentation every Foveon picture must be destroyed by antialiasing. However, as he looks at the pictures and they do look nice most of the time. People say they get them straight out of SPP without much tweaking. His latest theory is, this kind of aliasing does improve the picture, adding detail which is unreal and should not be there! Well, to close the circle to the audio theme, why should we care? Similar is said about the tube amps which are preferred by their fans because of their sound.

Without AA filter Sigma is on par with the most famous high end camera manufacturers as Leica, Hasselblad and others... Beside all weaknesses which are assigned to Sigma technology, let's be proud of this fact, and considering the rude inventation of Hasselblad to move their sensor for one pixel in each direction to emulate a full color sensor for still pictures, which is only possible without AA filter, welcome to our world!

Best regards
Wolfgang
 
Oversampling in audio has another reason and without knowing the facts (cost and effort reasons, marketed as a superior improvement to the crowd) should not be used as an argument here...
My opinion is that it can. Not without care though.
Whether an antiliasing filter is needed or not, everybody can decide by himself on hard facts, much more accurate than anybody else can do for him: Look at the picture, if there are no disturbing artefacts an antialiasing filter was not needed, if there are, the lack of an antialiasing filter has impaired the picture.
Thats too simplistic IMHO. You can look at an image you really like - and then someone comes along and makes it even better - and you like it even more. That you are not disturbed by artifacts in an image does not mean that the image cannot be improved.
The observation of many people here is, that the AA filter has an impact on the majority of the Bayer pictures, and they dislike it.
There are several reasons for this:

1. The AA filter needs to be too strong as it has to limit the spatial frequency for the blue and red detectors.

2. The LiNb filters only makes a double image - thats its entire filtering achievement. Thats not by far a good AA filter.

3. The optical properties for a LInB slab is sub par.

An oversampled sensor with a soft optical pre-filter and a steep digital post-filter would result in an optimal image. No such device exist. So - you cannot decide if you like it or not by looking at images.
Milling down the AA filter on Bayer cameras gives better results for some enthusiasts. In the professional field the Hasselblads etc. are delivered w/o AA filter, if required it can be used as an option. The Leica Digital Modul R, the Leica M series as well as the Leica S2 does not utilize an AA filter.
Yes, but see above.
One of the main reasons to use AA filters with Bayer cameras is, that the red and blue channel does degrade already at lower frequencies compared to the green channel, there are 2x more green pixels available than red or blue. This can become visible as very distractive color aliasing. The Foveon sensor has all color information for every pixel and therefore does not show this effect. Even if aliasing occurs, there is no color aliasing on Foveon pictures. Color aliasing is ways more disturbing.
As I stated above - one reason why AA filter dgrades the image so much for Bayer cameras is that they have a sparser sampling of red and blue. If you sample all color in the same spot - then the AA filter can be weaker.
For the same picture and lens, the aliasing artefacts will be higher for a sensor with less pixels.
Yes
Contrary to random noise characterised by the same power for all frequencies real life pictures are bandlimited and carry less power for the higher frequencies.
Yes - due to the natural AA filter called unsharpness.
For our eye the residual errors are oven more than acceptable.
Thats a matter of opinion. Do you like the rope artifacts in Foveon images?
... lots of nice info ...
Yes ... the eye and the ear works differently. Phase is very important for the eye. Frequency is very important to the ear. etc etc

But ... I do think it is premature to say that you cannot improve images by oversampling and AA filtering. I am quite sure you can ... as soon as you find out how you can do it cheap enough.

--
Roland

support http://www.openraw.org/

X3F tools : http://www.proxel.se/x3f.html
 
...continued (more argumenting - that I partly agree with and partly not)
Without AA filter Sigma is on par with the most famous high end camera manufacturers as Leica, Hasselblad and others... Beside all weaknesses which are assigned to Sigma technology, let's be proud of this fact, and considering the rude inventation of Hasselblad to move their sensor for one pixel in each direction to emulate a full color sensor for still pictures, which is only possible without AA filter, welcome to our world!
Thats not entirely true. If you use a weaker AA filter that is designed for a full color 50 Mpixel camera - then it would work very well.

Still - the LiNb AA filters are bad. So ... it might be a bad idea to use one.

But if the camera was say 500 Mpixels - then you could probably use the lens as pre-AA-filter and then use the printer as post-AA-filter.

--
Roland

support http://www.openraw.org/

X3F tools : http://www.proxel.se/x3f.html
 
I'm right down the road a bit... I'm right off the Fletcher St. Exit of 75 South of you in Tampa (Temple Terrace actually).
 
Tom, I've used the D70 and the D100 myself... While I liked some of the features of the D70 the D100 was my first DSLR and I learned how to use that camera to a T... Well, when I tried to do the same things with the D70 It took awhile to get back to where I was with the D100 but eventually I got close...

Basically, when you get the Sigma understand that it is a different world and will take some time to get used to... Also, note that custom white balance and good exposure go a long way with the Sigma line of cameras but take heart because their Metering is actually very good at protecting highlights without under exposing in most practical situations.
Ha ! Waiting for new DP to arrive tomorrow. Shoot lots with D70 and Rick Decker's image posts (from same sites) make mine look like trash.

Was going to try new Sigma wide zoom, but let's see if DP makes me a hero ...

Take care,
Tom B
 
Is that Fuji put a ton of effort into making Bayer work for them... Better Contrast, Better DR, Better Pixel Level Sharpness and More Vibrant Colors and yet that's the sensor that is no longer being made in a DSLR...

I like Nikon cameras but I'm disappointed that it requires Primes or really expensive glass to get good detail where as the Foveon makes mid range lenses look plenty sharp enough.
 
The Kodak's removed the AA Filter and used 14 MP Bayer, guess what... They had Moire like Crazy because of the Bayer Interpolation so while theoretically a 15 MP camera wouldn't have as much aliasing it would have other issues to contend with...

Basically, I believe if you can get a 10 MP Foveon (10 MP per layer) without AA you won't have either Aliasing or Moire which is why this technology still has more promise.
 
The Kodak's removed the AA Filter and used 14 MP Bayer, guess what... They had Moire like Crazy because of the Bayer Interpolation so while theoretically a 15 MP camera wouldn't have as much aliasing it would have other issues to contend with...

Basically, I believe if you can get a 10 MP Foveon (10 MP per layer) without AA you won't have either Aliasing or Moire which is why this technology still has more promise.
Without running any simulations, or doing the brain exercise, believing easily leads to wrong conclusions. The first mistake is to set 14MP Bayer == 10MP Foveon. If there is a reason for Moire from a certain no. of pixels an infinitive number of pixels will alias or moire in all situations, but it won't.

The optical conversion process is linear for resolution. This means, if there is aliasing or Moire for a high pixel sensor it is possible to emulate this situation with a small pixel sensor by using magnification. E.g., if the FF Bayer sensor @ 14 MP shows crazy Moire with a 50mm lens use a SD14 for the same picture at the same position with a focal length divided by the crop factor of 1.7 and multiplied with the correction factor of 14MP/4.7MP. This will show you will need a lens of the focal length of 88mm or more for the Foveon sensor. We don't have that crazy aliasing and Moire on our pictures, and this proofes 'seeing is believing' ;)

BR
wolfgang
 
As others have said much of the time it's not required so it's better only to apply it where artifacts appear.
 
Thanks for the very relevant D100 _ D70 comment .... it brings some practical reality to the situation.
Posting two DP1 images asking for help .... aaarrrrggggghhh !

Talk about D70 to D100 and back first switch from D70 to DP1 was a disaster !

Tom B
 
Isn't it doable in software? Certainly seems like it should be.
Nope - its not doable in software. Not if you want real AA filters. You can of course try to remove the effects of lack of AA filter with some heuristics or some very aggressive smoothing. But - to really remove AA artifacts you need to do it before sampling - i.e. before its detected by the sensor.

--
Roland

support http://www.openraw.org/

X3F tools : http://www.proxel.se/x3f.html
 
I don't know if you're agreeing with me or not...

Either way, the Aliasing would more than likely disappear with 10 MP in most situations assuming the lens was capable of resolving to that level (look at the Leica Lens that came with the Panasonic DSLR, that clearly had excellent enough resolving power to minimize aliasing)...

Now as for the Moire, often the color variant is the result of Bayer Interpolation and these instances would never be seen on the Foveon Based Images.
The Kodak's removed the AA Filter and used 14 MP Bayer, guess what... They had Moire like Crazy because of the Bayer Interpolation so while theoretically a 15 MP camera wouldn't have as much aliasing it would have other issues to contend with...

Basically, I believe if you can get a 10 MP Foveon (10 MP per layer) without AA you won't have either Aliasing or Moire which is why this technology still has more promise.
Without running any simulations, or doing the brain exercise, believing easily leads to wrong conclusions. The first mistake is to set 14MP Bayer == 10MP Foveon. If there is a reason for Moire from a certain no. of pixels an infinitive number of pixels will alias or moire in all situations, but it won't.

The optical conversion process is linear for resolution. This means, if there is aliasing or Moire for a high pixel sensor it is possible to emulate this situation with a small pixel sensor by using magnification. E.g., if the FF Bayer sensor @ 14 MP shows crazy Moire with a 50mm lens use a SD14 for the same picture at the same position with a focal length divided by the crop factor of 1.7 and multiplied with the correction factor of 14MP/4.7MP. This will show you will need a lens of the focal length of 88mm or more for the Foveon sensor. We don't have that crazy aliasing and Moire on our pictures, and this proofes 'seeing is believing' ;)

BR
wolfgang
 
Are you now a hero? ;)

Where are the pics man?
Thanks for the very relevant D100 _ D70 comment .... it brings some practical reality to the situation.
Posting two DP1 images asking for help .... aaarrrrggggghhh !

Talk about D70 to D100 and back first switch from D70 to DP1 was a disaster !

Tom B
 
... still adjusting, but the pics are 'usable'. Not a great loss since many were also taken with D70 and GX200. The key for me is the detail and depth I've seen in so many Sigma images. Even with terrible WB, that depth and detail is there to a lesser degree, but those were handheld as well. Here is first RAW shot from my back property in St George, UT looking East. Even handheld the detail has quickly reinforced the decision to add DP1 as the key hiking cam.



Tom B
 

Keyboard shortcuts

Back
Top