nikon D7100 - how will Canon respond?

Started Feb 21, 2013 | Discussions thread
bobn2
bobn2 Forum Pro • Posts: 53,614
Re: Hopefully with no AA filter on the 7D II...
3

noirdesir wrote:

bobn2 wrote:

HSway wrote:

noirdesir wrote:

bobn2 wrote:

Pastynator wrote:

Cameras with no AA filter will inherently have more detail in their images.

This is not true. Camaeras with no AA filter may give the ´╗┐appearance´╗┐ of more detail, but in fact what you see is artefacting, not detail. With a bit of luck, the artefacts might make a convincing representation of detail, but it still isn't detail.

Bob,

would you consider blur to be an artefact? If yes, one is trading one artefact for another. And since sharpening can partially compensate the effect of an AA filter and blurring can partially compensate the artefacts of discrete sampling (ie, with no AA filter), you are a bit more stuck on the blurry side with an AA filter and a bit more stuck on the artefact side without one.

Although I am not Bob I understand the point he is making. In a brief -

unlike the artefacts, you are never stuck on the "blurry side" which is the difference. And quite the opposite, by using your favourite technique to fine-tune this crucial aspect of the image representation, depending on your output intent, you take advantage of the given options. Whether one appreciates it or recognizes it as an advantage is another matter.

Crunchy rendering in the sky or of the water surfaces can be even more annoying. A layer mask and adjustment can help with this type of objects, though, often not completely and means an extra work.

OTOH, I see advantage of no AA filter with aps-c sensors in very slightly lesser need for sharpening which, despite a small difference, still results in less visible noise. That especially in crops and at high isos, but not only.

-- hide signature --

Thanks for standing in. Quite right, I would not consider blur to be an artifact. What blur is, is a filter - that is it reduces the amount of information, while aliasing creates false information. Blur can be restored by exchanging frequency response for dynamic range - that is by applying a filter that is complementary to the blur filter the frequency response can be restored, and the cost of amplifying the noise at high frequencies - this is what 'sharpening' does. The false information created by aliasing cannot however be removed, because there is nothing to distinguish it from real information, except a preconception of what reality ought to look like.

Take astrophotography, if the AA filter just manages to blur a distant star such that is no longer discernable, doesn't the image show something that is different than the reality due to the presence of the AA filter? An isn't showing something else than reality the definition of an artefact?

No, it isn't. This is a semantic game. If you wish to define 'artefact' as any variation from reality, then that is up to you, but that is not how the term is conventionally used in sampling theory. Play around with the definition, fine, but it doesn't change anything, just means we have to invent new terms.

I can fully go along with an answer that says that the showing something different than reality when reaching the resolution limits (of the sensor, a lens) is the most common case of showing something else than reality, such that humans have given it a separate name: resolution limitations.

Reality has resolution limitations, and in photography we are well used to resolution limitations. Aliasing artefacts are something different, in look and in what you can do about them. Different, hence different term. Clarity is not helped by conflating different things under the same term.

But that is only the introduction to my post. The main part is that I am trying, as carefully as possible in my wording, to describe the advantages and disadvantages of an AA filter. I fully agree that what most people see as 'accutance' (or whatever word they use) is to a large degree artefacts. But I also believe that (a) inverting the effect of an AA filter is an ill-posed problem (unless you had an absolutely noise-free signal) and (b) we don't know the precise AA function.

Every properly design sampled system requires an anti aliasing filter. The reason is that i you don't you produce information that is ambiguous - that is the same data can be produced by a number (potentially infinite) stimuli. In this case, the image projected on the sensor is the stimulus, and the signal produced by the sensor is the data. Once you have that situation, it is impossible to determine which of those stimuli is correct, choose the wrong one and you get artefacts. An unartifacted data set represents a complete, one to one description of the image, within the limits of the sampling theory.

In practical terms there is no imperative to be 'correct'. Just like many people prefer the sound of the distortion of a tube amp to the accuracy of a transistor amp, you may prefer the look of aliasing. Not a bad thing, but just accept it for what it is.

-- hide signature --

Bob

Post (hide subjects) Posted by
(unknown member)
ajy
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
tko
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
tko
MAC
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow