MG_BRN wrote: Better upgrade my Computer .. The difference is amazing
Actually ... before y'all get too excited ... better upgrade your lenses, since the DLO applies only to L lenses and a select few others like the 50 f/1.4. Third party lens, no hope.
Convolution is the mathematical term for applying an operator to data. A Gaussian blur is a symmetrical operator that takes data from each image point and adds a little bit of it to to neighbouring points. Think of passing a mound around the image over every point so that a little of the center of the mound rolls down the sides. We say that the image has been convolved with the Gaussian.
Len blurs are similar but more complexly shaped: they may have rings with depressions where a little is subtracted around the central peak and they may be assymetrical.
If you know the operator that was convolved with the image, you can deconvolve it by convolving the inverse operator. It is a little like division is the reverse of multiplication.
An optical device like a lens is a kind of analogue (non-digital) computer. It convolves an operator with the light coming in. A perfect pinhole (neglecting diffraction) would be an identity operator, i.e. no change (rotation of the image is not a significant change).
Lenses are not perfect operators but have aberrations and diffraction and distortions. If you test a lens very carefully (more rigorously than people can at home or in your average professional studio) you can determine the operator that corresponds to the lens. Then you can take an image produced by that lens and deconvolve it to get a nearly perfect image.
However (there is always a 'however'), lenses in the field depart from the ideal lens or the measured lens by small maladjustments and manufacturing deviations that are well within stringent quality controls but are simply due to the variability (stochastic nature) of any physical process. Also, lenses age and suffer wear and tear. So deconvolution with one set of measured data will never be perfect.
It seems that according to the pre-release instruction manual, at the present time, only the following Canon lenses are supported (though this may change before the official release of the next version of DPP, and it does not take into account future profiles Canon might decide to create):
EF 14/2.8L II USM
EF 24/1.4L II USM
EF 35/1.4L USM
EF 50/1.4 USM
EF 50/1.2L USM
EF 85/1.2L II USM
EF 300/2.8L IS II USM
EF 400/2.8L IS II USM
EF 500/4L IS II USM
EF 600/4L IS II USM
EF 16-35/2.8L USM, II USM (both versions)
EF 17-40/4L USM
EF 24-70/2.8L USM, II USM (both versions)
EF 24-105/4L IS USM
EF 28-300/3.5-5.6L IS USM
EF 70-200/2.8L IS USM, II USM (both versions)
EF 70-200/4L USM, IS USM (either with/without IS)
EF 70-300/4-5.6 IS USM [not the L?]
EF 100-400/4.5-5.6L IS USM
EF-S 10-22/3.5-4.5 USM
EF-S 15-85/3.5-5.6 IS
EF-S 17-55/2.8 IS USM
EF-S 17-85/4-5.6 IS USM
EF-S 18-200/3.5-5.6 IS
EF-S 18-135/3.5-5.6 IS
I think the range of supported lenses suggests that Canon chose designs that (1) they felt would benefit from this algorithm, and/or (2) are especially popular lenses, and/or (3) tend to be used by photographers who wish to extract the maximum performance out of the system.
I am, however, somewhat surprised that the macro lenses were not listed. The EF 100/2.8L macro IS, EF 180/3.5L, and MP-E 65/2.8 1-5x macro are frequently used stopped down for maximum DOF. And to some extent, I could also imagine DLO being useful for the TS-E lenses, so it's a shame we don't see those in the list, either. It's entirely conceivable that we might see more Canon lenses supported in the future, though.
As for third-party lenses, I think it would be impractical for Canon to spend the resources to measure a competitor's lens and create profiles. It may not even be feasible, since for all we know Canon could be using their own ray traced simulations of the spot diagrams of their lenses to compute the profiles.
As for me, I have 3 lenses in that list--4 if you count ones I've since parted with but from which I still have images.
RazorTM wrote:
From what I remember from reading the Luminous Landscape forums, this is called deconvolution sharpening, and basically yes, it uses the lenses' properties to work backwards to eliminate the effect of diffraction. Apparently, Lightroom does this to a small extent to eliminate the blur from the AA filter, but isn't extremely effective because it doesn't have a very good model for each lens like this new Canon system. When you move the Detail slider to 0 on the sharpening portion of the Develop module, you're using 0% deconvolution sharpening, and when you move it toward 100, you're using 100% deconvolution sharpening (0% USM sharpening).
The difficult thing with what Canon is trying to do is the point spread function (response of the system to a point source) of the lens + filter system is extremely complex and spread over a large number of pixels. The image you get out of the lens + filters system is the convolution of the point spread function and the image you would get in a perfect diffraction limited system. In an ideal diffraction limited system free of any aberrations the PSF is the Airy disk whose extent is no more than a pixel. In frequency space you can technically just divide the true image response by the PSF to get the ideal image response. In practice, finding the true PSF is impossible, so it must be approximated, either theoretically or experimentally measured, but it is non-linear and varies over the lens. The better the scheme the more computationally intensive it is, so don't be surprised if this takes some real grunt. Looks like Canon has done a very good job of undoing some of the effects of lens imperfections without introducing other artefacts.
As for downloading the software from the link in the OP, much as I'd like to try it out, I will not. I assume that doing so could/would constitute a willful violation of copyright, and perhaps other laws (e.g., trade secrets), and would subject an offender to serious civil and/or criminal liabilities should Canon decide to pursue the issue. Use your own judgment.
As for the technical issues, while we can't be sure what Canon is doing, I agree that it's most likely a deconvolution with a lens-specific, spatially, and probably focal-length- and aperture-, varying point-spread function.
Every real lens blurs an image a little by convolving an image with a point-spread function. This blurs/mixes light at each point in an image with light from nearby points. The result is an attenuation of higher frequencies (e.g., of sharp edges). That is, such a convolution is generally some form of low-pass filter.
What a deconvolution algorithm does in essence is to amplify attenuated high frequencies to restore high frequency details (e.g., sharp edges. However, along with these high frequency signals, these algorithms also amplify noise (as if you were cranking up the ISO).
If a high frequency signal is attenuated enough by the lens blurring, i.e., if it's diminished to the point that the resulting signal-to-noise ration is very low, then you'll end up with a lot of noise in the reconstructed signal. If the attenuation pushes the signal down to the level of the noise, no amount of recovery can restore it, and if you try you'll only get noise.
Deconvolution algorithms can also produce other artifacts, such as ringing at high frequency transitions (artificial undulations in intensity).
All this being said, it's still *wonderful* that Canon is seeing fit to provide lens-specific deconvolution operators. It's a bummer that we didn't have this capability sooner from the lens manufacturer.
Fred Miranda wrote:
It is indeed very impressive. However there is noise increase. Check out the sky areas.
Consider Canon's new DLO software to be a major step in camera manufactures becoming image software companies. More advanced versions will evolve and Nikon too shall evolve into this arena.
Adobe and other existing software companies are posted on Canon walls as Outlaws ...
Outlaws that capitalize on Canon camera coat tails.
In the future world imagined by Canon there will be no software companies infringing on images from their cameras.
This also goes for hardware like lenes, flashes, etc.
Canon allowing Adobe to use ... LMAO ... They are out for blood !!
I am now considering using DPP as a part of my workflow, in addition to LR but I am not looking forward to the additional time to do so . I need to figure out a way to streamline that process.
Stargazer 78, do you know whether this feature will work on TIF files or do they have to be RAW?
Monito wrote: If you know the operator that was convolved with the image, you can deconvolve it by convolving the inverse operator. It is a little like division is the reverse of multiplication.
This is not true, either in general, or specifically as applied to photographic images.
An easy demonstration: let me take an image and apply a simple, well-defined convolution to it, say, Gaussian blur. I will then give you the blurred image and ask you to deconvolute it. Will you be able to do it?
Just tried it on couple of holidays shots. The extra detail resolved is impressive! Not somethin that I've seen being done easily with sharpening tools available on Photoshop though.
However, it does introduce certain amounts of noise/artifacts on smooth graduated tone areas as pointed out. Which can be solved with a little Noise Reduction and masking of course as far as I tried. .
KaaX wrote:
An easy demonstration: let me take an image and apply a simple, well-defined convolution to it, say, Gaussian blur. I will then give you the blurred image and ask you to deconvolute it. Will you be able to do it?
The image alone is not sufficient. You need to know the operator as explained above. That would be the lens data detrmined by Canon. Even that is imperfect, as explained above.
Further, the deconvolving is senstive to noise, as explained above.
So Mr Stargazer, how do you think this will apply to astrophotos? Automatic flat-fielding? How does the 5D3 hold up for astro from a noise perspective? (or did you get hit with the new-gear curse)
Better upgrade my Computer .. The difference is amazing