If there’s one thing we should have learned this far into the era of digital imaging, it’s that nothing should really surprise us any more… and no development in technology – no matter how fantastical it may seem – is beyond the realms of possibility.
So, in my last column, under the title ‘Size Matters’, I considered what might be the implications of Canon’s 50 megapixels D-SLRs. Fifty seems to be a magic number in megapixels digital capture and it’s certainly been important in the digital medium format world recently as all the major players promote the various benefits of a 50 MP ‘645’ CMOS. Canon’s full-35mm sensor obviously has smaller pixels, but the new EOS 5DS have plenty of other attractions related to speed, operational efficiency, lens system and more. But… are smaller pixels really an issue any more or, for that matter, smaller sized sensors? Perhaps not.
Elsewhere in the current issue of ProPhoto (71#2) you can read our evaluation of Olympus’s new Mark II version of the E-M5. The original was the first OM-D series mirrorless camera and, subsequently, Olympus has been slowly but surely powering its way back into favour with both enthusiasts and pro-level photographers. The OM-D system is based on the Micro Four Thirds sensor format – the smallest of all the ‘mainstream’ sizes – with a maximum pixel count of ‘only’ 17 million. But Olympus has found an ingenious way around what might be considered the limitations of its sensor size and resolution – pixel shifting. The idea isn’t new – Hasselblad has been offering multi-shot versions of its capture systems for ages – but Olympus is the first to apply the concept to a ‘small format’ sensor. It’s a development of the sensor-shift method of image
stabilisation which Olympus has refined to enable sub-pixel amounts of displacement.
The end result is a 64 MP RAW image which is 100 MB in size. Obviously because multiple captures are involved, the prerequisite is that both the camera and the subject be rock-solid still, but it’s not really all that hard to see future developments overcoming this via a more sophisticated form of image stabilisation. Olympus itself is suggesting future OM-D
cameras will allow for high-res multi-capture shooting without a tripod.
As it happens, Ricoh is proposing a similar system for its planned full-35mm format
Pentax D-SLR (due later in the year), but its idea is not to increase the resolution, but to leverage some other image quality benefits that come from pixel shifting. Ricoh’s system uses four captures with a one-pixel shift in each direction (Olympus employs eight which includes four half-pixel shifts) so that, in the end, each pixel contains full (i.e. non-interpolated) RGB data. The benefits of this are the elimination of colour moiré effects and interpolation artefacts plus improvements to both sharpness and colour fidelity.
Right now, in its first iteration, pixel-shifting has limitations, but just look at how everything else that’s been introduced in digital cameras over the last few years has been subsequently refined and revised to work effectively. Today’s processors are more than up to the job of controlling, with sub-pixel precision and handling, the vast amounts of data involved. It’s likely then that, in the future, the physical characteristics of a sensor (i.e. the size and pixel counts) won’t matter – within reason – and you’ll be able to dial in higher resolutions as and when they’re needed. Olympus’s research has shown that there’s probably an optimum number of shifts per capture, but it points out that image quality is a product of many factors beyond just the sensor, including advanced image stabilisation systems.
If you thought we might have gone as far as we could with imaging sensor design and performance, in reality it looks like we might still only just be at the beginning.
Paul Burrows, Editor, ProPhoto