Most lenses tend to be expensive because the lens needs to be very complicated in order to correct for various optical effects (aberration, etc). Is it possible to build a sensor and processor combination which would fix defects produced by cheap lenses? I'm assuming the processor knows exactly the parameters of the len's defects.


Not only is it possible, but it's becoming commonplace. The micro-four-thirds system makes extensive use of it, and some compact cameras now do too. (I imagine that if they don't yet, most super-zooms will within a few years.) Digital Photography Review has a good article on this at http://www.dpreview.com/articles/5653763779/a-distorted-view-in-camera-distortion-correction , and it's worth reading even if you don't like that site's reviews in general.

Correcting distortion and lateral chromatic aberration is an option in my Pentax K-7, but although it doesn't impact shot-to-shot time (unless you fill the buffer), it takes a few extra seconds per image.

  • mattdm, Your line about dpreview;"and it's worth reading even if you don't like that site's reviews in general.", seems to imply you might not be a big fan of their reviews. If so, why not? Just curious. – Ken Jan 17 '11 at 14:06
  • Well, everyone always accuses them of bias whenever they print a review that says anything slightly negative of "their" brand of camera. But it's not that. The problem is: although they are a camera-oriented site, they are ultimately gadget-focused. The reviews focus on technical measurements, giving an aura of objectiveness. But there's problems.... – mattdm Jan 17 '11 at 14:16
  • First, the differences in these measurements in the objective sense is very small within a given generation of cameras, and the comparative reviews exaggerate them. – mattdm Jan 17 '11 at 14:17
  • Second, there are subjective differences in handling and use which are actually very important to making a nice, polished camera, and since those are "soft", they barely get talked about. The ergonomics and controls of the camera are described but not reviewed. – mattdm Jan 17 '11 at 14:18
  • Third, there are important measurable differences which don't get measured because they're not in the site's test suite. For example, color response gets a chart buried in the middle of the review, but not the charts-and graphs treatment. Or: spot-on auto white balance is a time-saver even when working in RAW, but at best that gets a line like "Poor auto-wb, but that's normal" in the "cons" section at the end. – mattdm Jan 17 '11 at 14:23

There are a limited set of defects that can be corrected in software, lateral chromatic aberration, yes, but not longitudinal chromatic aberration. Lateral CA results in the component colours of light being displaced radially across the sensor. This can be corrected by simply warping each colour channel slightly differently. Longitudinal CA causes out of focus areas near strong lights to be tinted either green of magenta depending on whether they are in front of behind the plane of focus. This is much harder to correct as you need to know 3D information about the scene.

Some cameras are now offering lens corrections based on a database of corrections tailored to each lens. I think these are limited to barrel distortion, lateral CA and vignetting.

You'd think the manufacturers would be the best source of this sort of information but I would image crowd sourcing the parameters on the net would be more effective as there are some very dedicated people around taking that sort of measurement.

  • longitunal CA can be corrected in lytro-like cameras, together with other focusing-related errors, like field curvature. this allows for greatly simplified lens. light field cameras are exotic animals, but it's nice how they show us what new things could someday appear in the mainstream. – szulat Nov 16 '14 at 1:15

If it is possible in post-processing, then technically it can be done in-camera. Some defects are pretty easy to fix - like geometry defects, but other are very hard, if not impossible, to fix in software (soft focus [yes, you can to some extent generate a sharper image, but this is not really reversing the lens imperfection], chromatic aberrations, ghost images or flare due to uncoated elements, etc.).


Go look up "Schmidt Camera" which is a type of telescope for photographing wide areas of the sky. The optics create a spherical abberation, so the telescope curves the film to match. A schmidt camera can produce an image of the sky with pinpoint stars at f1.5, which is rather impressive.


There are a certain class, like things that can be done in software, but there isn't a magic piece of glass that could be added to a camera to make it perform better.

One thing that might be possible is to have lenses built in 2 part systems. The part closest to the camera might handle the more delicate lens corrections, and the further away part give the right range. Still, something tells me that while this might be possible, it would be more of a pain then it's really worth. Hmmm...

  • I was thinking about in camera image processing rather than correction using glass. It's the cost of all those correction lenses which we are trying to avoid. – Ken Jan 17 '11 at 13:47
  • Right, but the right set of correction lenses might be a one off expense, as compared to using the same/similar lenses for everything. – PearsonArtPhoto Jan 17 '11 at 14:09

I don't know how many 'cheap' lenses are catalogued, but DXO does a very good job correcting lense-induced faults, in my experience.

As a specific example, I used it to correct pincushioning on architectural photos taken with a Canon EF 28-300 f/3.5-5.6L IS USM at 28mm and the results were impressive (curved walls at the edge became perfectly straight).

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.