DXO Analyser & Optics Pro

Discussion in 'Digital Darkroom' started by og, Feb 3, 2004.

  1. og

    og

    This information has not arrived on PN yet, so I am the first to talk about it... Cool :)
    Do Labs has created 2 softwares:
    - DXO Analyser is using a target to measure distortion, chromatic aberration, blur, vignetting and noise of any Capture device (Lens+Sensor...). It can build up a profile for all lenses+bodies (taking into account focal, aperture, distance...) to record those measurements.
    - DXO Optics Pro can take advantage of existing generic profiles (lens+body) in order to correct automatically those major optical faults in all your pictures.
    => It will really improve the performance (quality) of your lenses.
    Have a look at the website, the technique parts and the gallery...
    Ok... My point of view is that this piece of software (DXO Optics Pro) is really impressive, and may be the most usefull part of your digital darkroom... It may also have important consequences on the design of future lenses (consider trade-off on quality VS weight, size, cost...) .
    For those who are interested on the business side of the story, just consider that cheapy lenses of cameraphones will not show distortion, chromatic aberration, blur or vignetting when fitted with this technology (this is probably the huge part of the business... :).
    I really think that this is big news... any comment to share?
    Olivier
     
  2. Olivier, please stay with both feet on the ground about this.
    Just a simple example:
    If, e.g. due to a lens fault, your captured image contains a block of e.g. 2x2 white pixels, this software is never able to give you the REAL pixels corresponding to the scene you wanted to capture.
    Software can interpolate or extrapolate the actually captured information to create (i.e. make up) new pixels, but it can NOT replace what just isn't there.
    The only advantage I see, is that with this software you can sharpen/straighten/colour-adjust etc. your captured image "localised", i.e. corresponding to your profiled lens/sensor. However, some pixels will be compressed, others expanded in the process.
    Just to summarize my point of view:
    Software can never replace what is not captured by your lens/sensor system.
     
  3. og

    og

    Jos. I agree with you, and your post is exactly what was needed to counter-balance my 'WOW' optimism right now. Thank you :)<br>
    Also, my statement "It will really improve the performance (quality) of your lenses" is incorrect. I should have written "It will really improve the quality of your pictures".<p>

    But still... WOW! :)
     
  4. The expectations of DXO are skyrocketing. Ever the optimist I'm hoping it's something very special, however Genuine Fractals also seemed to promise the earth, but delivered something less than a full blooded miracle.
     
  5. "Have a look at the website, the technique parts and the gallery..."

    Am I missing something? I found the gallery part of the website, but only images from camera equipped mobile phones. On the home page it talks about improvements from good lenses, and huge improvements from poor lenses. Well you don't get poorer than a mobile phone but all I saw was nasty versions of nasty originals. I was hoping for more as I've got a few Canon D1s wide angle shots with colour fringing, and some vignetting from tilt and shift lenses, all of which would benefit from a fully performing version of this software.
     
  6. This isn't so revolutionary. These sorts of corrections have been used on a daily basic for decades in other fields that use digital image processing e.g. medical, military, astrophysics, robotics.
    I have no knowledge of this product but basically the software almost certainly combines the effects of multiple filters, levels, blending etc. already available on Photoshop. The benefit of this product would not be that it can actually do anything not already possible in Photoshop, rather that if implemented correctly, it might very well ease the process of image correction.
    Several issues:
    1. barrel/pincushon distortion: The "Liquify" filter allows image correction by a mesh.
    2. low image contrast/flare: levels/curves
    3. low image sharpness: unsharp mask
    4. chromatic aberration: unsharp mask to individual channels
    Many of these operations can be classified under the heading of: deconvolution where the image and targets are transformed to the Fourier space. In this case, a Target T and an imaged Target Ti are divided to form a correction matrix T/Ti = Tc, and then an image I is corrected by I*Tc = Ic, and then transformed back to the cartesian space.
    A number of custom convolution kernels can be applied to an image in Photoshop via the "Custom" filter (you need to know what numbers to type in though).
    In any case if these functions can be automated, it would save alot of time. On the other hand there are limits to the ability to recover information via such transforms. Don't expect to use a disposable plastic lens and get the same quality images as if using a Summicron 50, nor should you expect to be able to transform an out of focus picture of your foot into a professional quality portrait -- for those actions you'd need to paint the picture pixel by pixel :)
     
  7. og

    og

    Gary,
    I looked closely at this example (from 'Original Image' to 'DXO-optics correction'(only): chromatic aberations & distortions are very well corrected.
    On the other side, I am not very happy with the 'LocalContrast correction', which is too harsh for my liking. This effect may indeed lower the perceived quality of other samples (I guess that they have to work on this one...). This sample also suffers from jpeg artefacts, and the increased sharpness may have emphasized it => the RAW/16bits module will be a must, I guess (or at least better jpeg files).
    Too see a better example about vignetting correction, check this one (although 'LocalContrast' is really not welcome once again...).
    They will probably show more interesting examples after PMA.
    The french magazine 'Chasseur d'Images' has bought the DXO Analyser and their next releases will feature several tests of lense+body. In particular, they have come up with extremely poor ratings for the Canon 1Ds + 16-35mm f/2.8. So maybe Do Labs will try to demonstrate its new product on the 1Ds+16-35... that would be very interesting!
    Jonathan,
    "The benefit of this product would not be that it can actually do anything not already possible in Photoshop, rather that if implemented correctly, it might very well ease the process of image correction." + "(you need to know what numbers to type in though)"
    I guess that this is where their actual measurements of all lens+body combinations (+focal length, aperture, focus distance, etc...) is invaluable to come up with the good 'numbers to type in' :)
    There is nothing new... except that if they have come up with a way to automatize and get good results for a reasonable price... this will indeed be new.
    Olivier
    ...still optimistic :)
     
  8. Did anyone notice that the "per-site" license/price of the ANALYSER 1.0 software is:

    $18,750.00 US?

    That's right - TWO figures to the left of the ",". ONE-EIGHT thousand.

    Plus $500 for the calibrated "target".


    http://www.luminous-landscape.com/about/dxo-press.shtml

    I doubt many photographers will be buying a copy to profile their own cameras/lenses. You will get the "canned" profiles camera-makers create for their own cameras (probably eventually as part of the RAW-capture software).

    It IS an ingenious idea, though - sort of "lens profiling" to match the monitor, scanner and printer profiling done by color management suites using IT8 targets.
     
  9. This technology (although very cool) is not "revolutionary". You can read about it under general name of Adaptive Optics. e.g. see:
    to name just a couple of sites that pop up under google. I would imagine that any sophisticated lens manufactorer e.g. Leica, Zeiss, Schneider, Canon, Nikon, would already use such mathematical technology to optimize their lenses. Of course when any of these folks let us know that they are using this product, I would be impressed :)
     
  10. Like Olivier, I am an optimist. Isn't it great that consumers are going to be able to use technology that was available only to the military? For example, I do not know how to correct out of focus images using only Photoshop. Does anybody else?
    In a similar vein, this also looks interesting.
     
  11. og

    og

    Update: Imaging-Resource has been quite impressed by the results of DxO Optics Pro so far. See samples here.
     

Share This Page