Aliasing: the Moiré pattern
Large format sensors* with high resolution lenses suffer
one problem: Aliasing (which appears in an image as a moiré pattern).
This can occur anywhere in an image where the frequency of detail is higher
than the sensor's Nyquist frequency. The Nyquist frequency defined as
the highest spatial frequency where the CCD can still faithfully record
image detail. Beyond the Nyquist frequency aliasing occurs.
A special type of filter called an 'Anti-Alias Filter' can be placed in front of such a sensor which essentially filters out any frequencies above the Nyquist frequency. Digital SLR's such as the Nikon D1/D1x/D1H, Canon EOS-D30 and Fujifilm S1 Pro all feature anti-alias filters on a piece of glass attached directly to the surface of the sensor. Kodak's DCS series cameras have always had a separate anti-alias filter which is mounted just behind the lens mount, in front of the mirror.
The disadvantage of an anti-alias filter is that it does soften the image somewhat. Many Pro's shoot without an anti-alias filter (instead choosing to use a simple Infrared filter) and deal with moiré in software later (see 'Moiré removal' below).
The DCS 760 doesn't ship with an anti-alias filter (it's an option), instead Kodak have chosen to supply it with an IR filter. Clearly the more pixels you have on your sensor the less of a problem aliasing becomes, and in our observation of the DSC 760 this is very true. Thinking about a six megapixel sensor logically you need a very sharp lens which is capable of generating very high resolution on the CCD's surface AND you need a picture detail which will generate frequencies above the sensor's Nyquist frequency. This WASN'T true of the lesser 2 megapixel sensor (found on the DCS520/620/620x) which would alias at the first hint of detail.
In our experience of shooting the DCS 760 day-to-day there were very few images were aliasing was visible or ever a problem which would 'ruin' an image. Where it did occur there is always moiré pattern removal in the DCS Photo Desk application (or using third party tools).
* Actually it's more to do with the actual size of each pixel on the sensor surface (cellsize)
As mentioned above the latest release of DCS Photo Desk (1.2.0b for this review) now includes moiré pattern removal. Below we'll be comparing it to the very popular third party tool Quantum Mechanic Pro from Camerabits (click here for a snapshot of QM Pro). All images were processed as TIFF and converted to JPEG later.
|Original image / 1,527 KB JPEG|
|DCS Photo Desk v1.2.0b Moiré removal enabled / 1,320 KB JPEG|
|Original image (top) run through Quantum Mechanic Pro / 1,351 KB JPEG|
|Original image / 1,869 KB JPEG|
|DCS Photo Desk v1.2.0b Moiré removal enabled / 1,632 KB JPEG|
|Original image (top) run through Quantum Mechanic Pro / 1,571 KB JPEG|
Both Photo Desk and Quantum Mechanic Pro do a good job at removing the worse of the bright 'traffic light' moiré spots, and they also cope with the majority of visible diagonal moiré, though it's clear that there is still some visible in the second image.
Photo Desk has three different levels of sharpening. I was a little disappointed about how harsh the Medium sharpening setting appears to be, there's also a serious lack of an Unsharp Mask option - something Nikon already have in their Nikon Capture application. All images were processed as TIFF and converted to JPEG later.
|Sharpening: None / 1,290 KB JPEG|
|Sharpening: Low / 1,427 KB JPEG|
|Sharpening: Medium / 1,643 KB JPEG|
|Sharpening: High / 2,025 KB JPEG|
|Sharpening: None + Photoshop: Unsharp Mask 190%,
Radius 0.5, Threshold 0
1,712 KB JPEG
I personally settled for Sharpening: Low for quick results and Sharpening: None + Photoshop Unsharp Mask (the last crop in this series) for the best results.