PetaPixel’s title Researchers Develop Method for Getting High-Quality Photos from Crappy Lenses is certainly more down to earth than the original High-Quality Computational Imaging Through Simple Lenses to which it refers, and unless you a more of a mathematician than me you will find parts of the latter hard going. But it is worth going to the original, if mainly for the short video and the roll-over images showing how well the method works.
I took one of the sample ‘before’ images from their simple lens into Photoshop and tried my best to improve it. With a couple of passes through my normal sharpening plug-in, FocalBlade I could make significant improvements to image sharpness (I have an old version of the plugin, and used first DeBlur Pro mode, then Selective Sharpen Pro) without gaining too much noise in smooth surfaces, but was very clearly left with considerable colour fringing when viewed at 1:1, and could think of no way to significantly reduce this in Photoshop. This is a problem that the researchers “novel cross-channel term ” is designed to resolve.
The paper seems to be another chapter in a story we have already seen beginning where lens designers can rely on software to minimise some lens defects. Many cameras already incorporate this, particularly for their jpeg images, which can be corrected for vignetting, distortion and chromatic aberrations, and ‘post’ software such as Lightroom can provide similar functions for interchangeable lenses, automatically if it has a lens profile. For most of us such profile corrections have become a part of our normal working practice, applied automatically by the software, both for cheap and expensive lenses.
It’s hard to know how much of a breakthrough the work of the team at British Columbia University is, particularly as the point spread functions they use differ at different subject distances. While this may not be too much of a problem with cheaper compact cameras (or phones) there may also be a problem in achieving the necessary computational power in these devices – though what looks like some pretty tricky maths to me may not be too difficult to a chip – who knows?
I’d certainly love to have simpler, lighter and cheaper lenses for the Nikon. Particularly with the lower back pain I’m suffering at the moment, lighter would be good! The 16-35mm f4 is a great lens, hard at times to believe how sharp it is (after a little help from Lightroom) but at around 680 g or 1.5 lbs it isn’t light hung on a camera body around the neck. And at around 5 inches long, plus a lens hood, with a 77mm filter, it is big. Good though the performance is, it still improves in Lightroom, which does a great job of removing distortion, chromatic aberration and vignetting.
This doesn’t quite all come for free, as the distortion removal does inflict a slight crop around the edges of the image, particularly noticeable at 16mm. Sometimes too, 16mm images can look better with the distortion, as when corrected to pure rectilinear there is actually more stretching at those edges. Round objects are actually closer to round with the distortion, and at 16mm most images actually look more natural without distortion correction – by setting the slider in the Profile sub-panel to zero. It would probably be a good idea to make this the default setting for this lens unless you are an architectural photographer. Correcting distortions isn’t always a good idea with wide angle lenses.
Lighter cheaper lenses don’t always mean poor quality. The Nikon 18-105 is a pretty snappy lens and a little under a pound in weight as well as smaller in very dimension than the 16-35. As Photozone remarks, it has rather high distortion and chromatic aberration, but again with a little help from Lightroom the results are generally hard to fault. Around a quarter the price of the 16-35mm, the biggest difference from the professional lens is in build quality. I can’t remember if the current lens I’m using is my third or fourth, though I did drop one of them. The low price means that repairs are seldom economic.