AGeoJO wrote:
Ouch, Canon wants to make sure that DPP is the choice of application, I guess. Thanks for clarifying that.
By any chance, do you know whether the modified RAW file can then be converted to DNG? I know it can be converted to TIF but at triple the size, the converted TIF file would be huge. At least, DNG files can still be processed using LR.
no it can't. At this time no one except for Canon can decode corrected CR2 files.
I wonder how much worse the DLO will perform on bodies that don't report the focus distance in the exif i.e. 5D classic? I known Canon restored that feature in the 5DII.
Some aberrations are focus distance dependent.
Mike
speedmaster20d wrote:
no it can't. At this time no one except for Canon can decode corrected CR2 files.
To be more precise, I thnk that Adobe could decode them, just not have access to the corrections. In other words, I expect the added correction information will not invalidate the basic CR2 info. Any definitive information on this clearer statement of the issue? i.e. by experiment?
In the medium run, it might be that dcRaw might gain access to the info by reverse engineering. It depends how obscure the coding Canon has used. Since dcRaw is the basis of ACR, Adobe might be able to get into the corrected images in the future.
In the long run, somebody like DxOmark could make the measurements of many lens and camera combinations and sell the point spread function information to deconvolution engine builders ... such as potentially Adobe.
Monito wrote:
In the long run, somebody like DxOmark could make the measurements of many lens and camera combinations and sell the point spread function information to deconvolution engine builders ... such as potentially Adobe.
DXO optics pro already does deconvolution and oterh aberration corrections. DLO a potential hit to them though they support 3rd party lenses and many more bodies Canon and non Canon. They don't have an inverse function for the AA filter though.
Michael Gordon wrote:
I wonder how much worse the DLO will perform on bodies that don't report the focus distance in the exif i.e. 5D classic? I known Canon restored that feature in the 5DII.
Some aberrations are focus distance dependent.
Mike
All Canon cameras after EOS D30 (not 30D) report focus distance and all of them will support DLO.
Nope, Canon stopped for awhile for unclear reaons. I am reasonably sure 40D and 5Dclassic (latter is certain) did NOT report focus distance in exif. Focus distance had to be put in manually for DXO optics pro to give optimal result. (D30 actully had focus distance in exif--go figure.)
DPP apparantly did not use the focus distance info so perhaps they took it out until Canon software required it--that is just a guess.
Shots taken with the 5D must have had the body loaded with firmware version 1.1.1.
Shots taken with the 30D must have had the body loaded with firmware version 1.0.6.
When focus distance is not reported in the EXIF, loading the profile is still supported (for lenses in the list), but the focus distance slider is set to infinity focus with a warning [!] symbol. The user may manually adjust this slider to achieve the desired result. The adjustment of this slider does not affect a similar slider in the existing lens aberration control panel.
wickerprints wrote:
Look, just stop explaining yourself to KaaX. He obviously doesn't understand signal processing, made an uninformed statement about deconvolution, and is now trying to insist that he was right.
Stargazer78, thanks so much for sharing the information and especially the download link! It is a nice new feature in DPP. Since I am using DPP exclusively to convert my RAW files, I will certainly make use of this feature. Just downloaded the lens infos I needed and tested it with some of my photos. Now I can nicely make use of my 8x i7 processor PC
After all my posts discussing the application of deconvolution to gaussian blur, I would like to point out this interesting (but perhaps disheartening) fact:
Point-spread deconvolution will not work to fix blur due to small-aperture diffraction.
Gaussian blur "leaks" enough high frequency information into the image that it can be boosted and reconstructed (indeed, a gaussian is a very special extremum case that gives "the most blur for the least loss of high frequency detail"). The Airy disc point spread function for diffraction from a circular aperture (and, more generally, the point-spread for diffraction from any shape of aperture) has the opposite property: it is the result of putting a hard cut on and zeroing out all detail frequencies outside a certain range. As a result, there is no higher frequency information "hidden" in the blur waiting to be boosted.
(Fun optics fact: the shape of your lens aperture is also the shape of the "window" in frequency space --- lower frequency patterns pass through, but, as the aperture gets smaller, more and more high frequency information is blocked out, producing a blurrier diffracted image.)
Deconvolution methods can still work on point spreads due to reasons other than diffraction (which limit sharpness at wider apertures), which probably do not entirely zero out all higher frequency data.
Fred Miranda wrote:
It would be really helpful if we could batch our images through DLO alone and have a resulting RAW file which could then be edited in LR or ACR.
Actually, to a certain degree you can batch the DLO. I tried.
You could try this:
- Copy an extra set of RAW files you want to be DLO processed and put in a separate directory.
- Open the directory in DPP
- Apply DLO on one of the images
- Right-click and "Copy Recipe to Clipboard"
- Select all the images that have not been applied yet, right click and "Paste Recipe"
DLO will also be applied to all the images. Granted these images must be sorted by the same lens I think. Not sure if you can mix and match lenses in the same batch. Haven't tried that.
- Meanwhile, if you have like 100 of them, go cook dinner or lunch and watch a little TV while the computer churns. My AMD quad-core 3.6ghz, 8gb of RAM takes about 15-25secs for an average 5DII RAW, to give some perspective.
kevindar wrote:
my i7 2600K 3.4 ghz, takes about 5-7 seconds per 5d2 file. however, I am not seeing a huge difference.
I don't have my files with me at work, perhaps can share some of the comparisons later when I get home.
Its quite impressive for some images and so-so for others. But at 100% view of all those I tried, its definitely very visible. Its very, very similar to what the OP demonstrated on my side of things too. Just whether it worth the effort to you.
I would say for people doing landscapes and maybe birding... its a game changer !
I did it, and was more patient with the previews. yes you are right, its very very visible, esp for landscape. for birding I did not see as much. seems like you have to push the slider all the way to 100 to see the biggest difference. Also, the biggest difference I noticed towards the edge/corner of the frame with my 16-35II image at f11.
For best comparisons, copy and past a RAW file in the same directory. Process with DLO on one of them. I find the DLO slider at 50% and 75% is the best as there are some artifacts with 100%.
Then open up both files in DPP and view 100% side by side. The differences are more evident when you do that versus the preview window.
mpmendenhall wrote:
After all my posts discussing the application of deconvolution to gaussian blur, I would like to point out this interesting (but perhaps disheartening) fact:
Point-spread deconvolution will not work to fix blur due to small-aperture diffraction.
Gaussian blur "leaks" enough high frequency information into the image that it can be boosted and reconstructed (indeed, a gaussian is a very special extremum case that gives "the most blur for the least loss of high frequency detail"). The Airy disc point spread function for diffraction from a circular aperture (and, more generally, the point-spread for diffraction from any shape of aperture) has the opposite property: it is the result of putting a hard cut on and zeroing out all detail frequencies outside a certain range. As a result, there is no higher frequency information "hidden" in the blur waiting to be boosted.
I appreciate your demonstration of gaussian deconvolution, but I question the assertion that the airy disc has a hard frequency limit in theory. The point spread function does not have a flat top. In practice though, the difficulteis with diffraction deconvolution may arise from a non-perfect aperture and a non-perfect airy disc.
. Thanks!
:-D