The K-5 IIs has created a lot of interest due to its lack of an AA-filter.
Many prospective buyers only seem to see the positive aspects (avoiding the additional blur caused by an AA-filter) but appear to not fully understand the disadvantages, claiming that artefacts such as moiré do not appear in practice and/or that any image issues are easy to remove.
In the interest of avoiding future disappointment, I'd like to clarify the pros and cons of a camera like the K-5 IIs that omits an AA-filter.
Two important messages upfront:
"Personally, I consider a sub-100MP camera w/o Bayer-AA filter technically flawed which is and must be sold with a disclaimer."
Statement made by Falk Lumo (falconeye). Falk is a Physicist and knows a thing or two about optics.
"For most photographers, we recommend using cameras with antialiasing filters. Antialiasing substantially reduces the risk and severity of demosaicing artefacts and moire. Apparent softness can be effectively overcome with Photo Ninja's excellent sharpening filter, which uses deblurring technology that can effectively compensate for antialiasing blur."
Statement made by the creators of the Photo Ninja RAW converter (of Noise Ninja fame).
Why do we have these two clear pro AA-filter positions?
Well, for the same reason DSLR manufacturers have been equipping their cameras with AA-filters for years without anyone forcing them to spend money on a precision part.
An AA-filter is an integral part of an imaging system using a sensor with a
Bayer colour filter array. Contrary to common belief, an AA-filter is not required to cut-off frequencies that would cause aliasing in the luminance channel. The fill-factor (the ratio of sensitive versus insensitive sensor area) is so high in modern sensors that luminance moiré does not play a significant role.
However, an AA-filter is nevertheless required because a sensor with a
Bayer colour filter array only records one colour component (Red, Green, or Blue) per sensel. Without the blurring caused by the AA-filter, a set of photons hitting one sensel would only have one colour component recorded for them. The role of the AA-filter is to distribute this set of photons so that it hits sensels of all kinds (Red, 2x Green, and Blue).
The following makes it clear that, strictly speaking, "AA-filter" is a misnomer for the birefringent plates that are put before imaging sensors. Watch the following demonstration of birefringence.
I hope it becomes clear that the AA-filter can be compared to a beam-splitter in a 3-sensor camera, that diverts the Red, Green, and Blue components of an image to dedicated image sensors. I furthermore hope that it becomes clear what the consequences are when no AA-filter is present:
- false colour: if single sensor sensels are excited in isolation, they will show up as colourful dots (red, green, or blue, depending on which type of sensel is excited), even though the image was monochromatic, for example.
- colour moiré: a consequence of the above, but much more visible as the mesh-like structure of moiré shows up on a much larger scale. Big image areas can be affected by (see the last example on the Photo Ninja page).
- demosaicing artefacts: The demosaicing algorithm that re-constructs full RGB information for each pixel from all the individual R, G, and B components in a RAW file can become confused by detail that should not have been recorded in the first place. The result are maze artefacts or similar artificial structures.
There are many reasons why moiré or other artefacts caused by removing an AA-filter do not always rear their ugly heads:
- The spatial frequencies in the scene are too low.
- The lens is wide open and acts as a low-pass filter.
- The f-ratio is high enough to cause sufficient diffraction.
- The lens is slightly defocused.
- There is motion blur from camera shake or subject movement.
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.
Contrary to what is often stated, moiré cannot easily be removed:
- It is always a manual process, requiring detailed attention. There is no way to "batch process" images, repairing moiré in an automated manner. You may leave the repair to a RAW converter that automatically addresses colour moiré but notice that one still can see where the moiré was. In other words, one may can easily suppress the false colour as such, but getting rid of the luminance pattern that comes with colour moiré isn't that easy at all. Furthermore, the converter may mess with information that it shouldn't mess with.
- Information has irrevocably been lost. Some invented information has to replace the ugly moiré patterns and this patching may or may not look good.
In contrast, the capture-sharpening required to remove the little loss in contrast caused by AA-filters, can be applied globally to an image and thus be made part of a default import setting (e.g. in Lightroom).
Only if the original image does not have pixel-level sharpness, due to any of the reasons listed above, then not using an AA-filter is actually of advantage. The reason is that the original blur, plus the blur added by the AA-filter adds to an overall blur that cannot be exactly reconstructed to the original detail.
Of course we are already seeing many comparisons allegedly demonstrating how the K-5 IIs blows the K-5 (II) out of the water in terms of recording detail. Note however, that few of these demonstrations have the rigour behind them that is required. A very prominent example is based on manual focus with focus confirmation by the camera (very inaccurate) and uses f/11 as the f-ratio (lots of details has been lost at such an aperture already).
Whenever you see a comparison, know that the version that was captured with an AA-filter always requires capture-sharpening. The Bayer colour filter matrix is ingenious in the sense that it records colour while recording spatial information at the same time. The blur that was necessary for the correct colour encoding, negatively affects the spatial aspect, which is why appropriate levels of capture-sharpening are always required. This is as essential as the correct white balance or gamma encoding. Do not let yourself be fooled by comparisons that show the same levels of sharpening applied to both AA-filter and AA-filterless versions.
I hope the above helps you to make an informed decision. If you expect almost all images you take to have some real world blur (slight defocus, camera shake, lens abberations, etc.) then the K-5 IIs may be for you.
If you aim at 100% pixel-level sharpness and use tripods, studio lighting, etc., to obtain the latter, the K-5 IIs may not be for you. Any detail beyond the colour resolution of your camera will cause artefacts and if these details are regular, they will cause highly visible and large colour moiré structures.
Arguably, most images will exhibit some slight blur and most scenes won't excite objectionable colour moiré. However, if colour moiré strikes, it can ruin your image and one could argue that this is worse than the slight negative effect of an AA-filter on your images. After all, most people will not exhibit their work in a form (e.g., poster size to be looked at from close distances) that would make the difference between AA-filter vs no AA-filter worthwhile. The issue with moiré is that moiré structures can be very large, certainly much bigger than the minute differences in pixel sharpness that are obtained by omitting an AA-filter.
The higher the number of MP, the less need there is for an AA-filter because real world blurring effects will play the role of a natural AA-filter. One such real-world effect is diffraction. When image sensors will record 100-200MP then even
diffraction at f/2.8 will be a sufficient source of blur. Lower f-ratios will cause other lens aberrations to dominate, so higher numbers of MP will not be necessary in order to avoid colour moiré. At 16MP, the K-5 IIs has probably too few MP in order to make the absence of an AA-filter truly practical.
My position is that the vast majority is much better served with a complete image forming system. Your mileage may vary.
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.
EDIT 2:
I concur with the
assessment by Imaging-Resource:
"Even then, a touch of unsharp masking on both images will often leave the K-5 II's result near-indistinguishable from that shot with the K-5 IIs.
For some purposes, that little extra sharpness might be worthwhile, but for my money it's simply not worth the risk of a hard-to-remove artifact in a once-in-a-lifetime shot. The slight reduction in sharpness is one I'm willing to make, for the peace of mind it brings."
You may check out some
real world K-5 IIs moiré examples (which is not to endorse the review whose scoring is broken in many ways).
Last edited by Class A; 06-08-2013 at 01:38 AM.