View Full Version here: : Canon Digital Photo Professional Digital Lens Optimizer
24-04-2012, 10:59 PM
This module is utterly fantastic.
Canon's Chuck Westfall on DLO:
I used it on a sample image from the "lowly" 17-40mm f/4L USM lens that I took in New Zealand last September.
The shot was taken at 1/6s f/16 at 17.0mm iso100 and is a crop from the extreme right of the frame.
The image is otherwise as-is out of camera with no processing applied.
The image on the left is without DLO, and, on the right, after DLO has been run on it.
DPP is the bomb.
Sounds fascinating Humayun. Some sort of convolution/deconv correcting for precisely determined characteristics of each lens?
24-04-2012, 11:39 PM
Yep, pretty much!
I'm very impressed by the way it handles the /entire/ image, however. The function is more prominent at the edges and in the centre, and, just massages itself elsewhere in the image, removing all chromatic aberration. DPP always had the ability to remove CA, but, it's now even better.
24-04-2012, 11:46 PM
Excellent Humayun. The advantage is obvious to see. I'll be giving it a run tomorrow.
24-04-2012, 11:50 PM
I for one don't use DPP as much as I should. It's so powerful.
24-04-2012, 11:50 PM
One trick to take note of is to set the initial sharpness setting to 0 before running DLO. Once you're satisfied with the results, then, return sharpness slowly, one click at a time, so that you don't introduce artefacts. With the example image above, I didn't even have to return sharpness -- it just came out that sharp after running DLO.
25-04-2012, 12:22 AM
Can't wait. I have many appropriate candidates in mind.
25-04-2012, 12:29 AM
Oh, and, by sharpness setting to 0, I meant on the first tab of DPP, not the second one with the histogram.
25-04-2012, 01:27 AM
I think this might be something half new (at least in this arena)
I am guessing (but dont know) they are using typical known and determined PSFs (Point Spread Functions) of lens aberrations to closely approximate the actual PSF of your particular lens and then run their algorithm to refine that PSF and then deconvolve.
Plus apply corrections to other sorts of aberrations and even possibly using different PSFs across the image to more accurately correct it.
One day some bright cookie will create something that can do this for all your lenses by imaging test targets and then accurately analysing properly and correcting. Meaning that an imperfect lens might be capable of taking a near perfect picture (after processing).
Quite possibly in the camera !
Which may in fact mean that lenses could be made much more cheaply as they dont have to be perfect, just be able to be analysed for their errors and then corrected.
Of course that might be expecting more form the technology than is currently possible, but I think in theory its possible.
BTW - While I can see the improvement in detail and contrast between the 2 versions of your image that the software has provided, I have to say that I actually prefer the first one !
It has that wonderful hand painted and aged toned look about it.
The second one just looks like a sharp landscape photo rather than a scene with mood and personality !
25-04-2012, 02:00 AM
Mate, I hear what you're saying.
I know that deconvolution and PSFs have been around for some time.
I do think it's great of Canon to put that functionality into their software, though. Each module is about 20-30 MB in size -- one module per lens. So, there's obviously something going on under the hood, more than just a simple algorithm.
My goal has always been to make photographs that resemble paintings. Church, Turner, Bierstadt, et. al., are my heroes. The only way to get the look is to photograph in the appropriate light. But, I still think the image post-DLO resembles a painting, just a bit sharper. :)
25-04-2012, 08:08 AM
Sounds great H. I'm using DPP more and more these days. How/where do we get a hold of the new version of DPP with this new DLO feature.
25-04-2012, 08:37 AM
That's a nice improvement H. As cameras get even more resolution and sharper that tool will become even more valuable as new cameras start showing up the weaknesses of existing lenses more and more.
25-04-2012, 11:56 AM
Interesting stuff H, DPP has always been a fantastic tool, not just for processing either, I found DPP would print correctly to my Epson R1800 while PS was all over the place and wouldn't produce consistent repeatable results.
I've been sorely disappointed with my 16-35 it's terribly soft at the edges, hopefully this update will show a big improvement.
I managed to find an update file here (http://www.canon.co.uk/Support/Consumer_Products/products/cameras/Digital_SLR/EOS_5D_Mark_III.aspx?DLtcmuri=tcm:1 4-923797&page=1&type=download)
25-04-2012, 12:00 PM
Thanks for reporting this H. Your example is very telling ... will have to give this a go soon - just need to find the right image to do it with.
25-04-2012, 12:21 PM
Sorry I didn't reply earlier, Greg. Phil's got the address for you.
Phil, I recall you've mentioned that DPP got the printing right for you on a couple of occasions. I still use Photoshop for printing because I run sharpening routines as well as brightening curves over the files before I print them.
One point to note is not to expect great results from DLO on very high ISO images. It introduces a lot of edge artefacts.
I would love it if people posted their before and after results in this thread.
It is an excellent module.
I will continue to extoll the virtues of using native proprietary software over and above reverse-engineered third party software.
Regardless of the fact that Canon provides an SDK for developers to write their own tools to decode the RAW files, none of them produce natural results that DPP does. I have always felt that we spend thousands of dollars on gear, why should third party software let you down? When I was uninformed I used to use Lightroom. I stopped using Lightroom because it kept aping the skin tones in my wedding images. People often say they use camera profiles for ACR to mimic DPP. Why should you have to? You've got a tool that does everything. If more work needs to be done on an image, there's an export to Photoshop menu item.
I, for one, expect to see on my computer screen what was on the back of the camera when I took the image. Not what Adobe et. al. thought the camera saw. DPP is the only tool that shows you exactly what your investment captured.
25-04-2012, 12:21 PM
From what I see in the new version there may be a need to add additional lens data info from the Lens tab/ Update in the toolkit window.
25-04-2012, 12:37 PM
What a difference H! Looks like a very handy tool.:thumbsup: I can see why you often head over to NZ with stunning views like that!:eyepop:
Thanks for the update link Phil.:thumbsup:
25-04-2012, 12:43 PM
Phil, yes, you need to go into the Lens tab and press Update. Then, tick your lenses and wait for them to download.
Each lens module is 20-30 MB in size and may take a little while to download.
Cheers, Daniel. I'll process that one properly, soon, and post it.
25-04-2012, 01:28 PM
After reading your info on DLO, I downloaded the update last night and yes, you need to download further each lense module.
Anyway, here is a sample from my job up in the Pilbara. The region of interest is to the extreme left of the original with text and fence wire to add complexity to the image.
For me this is fantastic as I often need to photograph detail wiring installations, model numbers etc.
Camera : 7D, ISO 200, 1/400th Firmware 1.2.8
Lense : 16-35mm F2.8L @ F5.6 & 20mm
Other Info : Polarising filter fitted. Shot in RAW (of course!)
DLO Setting : 50 (default when you select the checkbox)
Has made a noticeable difference and for me, this is a great module from an engineering use point of view. I also agree with the high ISO number impact.
Image 1 = Original unprocessed whatsoever, jpg export @ quality 6
Image 2 = DLO only, jpg export @ quality 6
25-04-2012, 01:45 PM
Chalk and cheese, my friend. Excellent result!
Though, the original looks like it didn't have any CA reduction applied in DPP prior to fixing it with DLO.
Not to worry, once you fix an image with DLO, the CA value on the tab returns to N/A I think.
It's interesting that you mention NOT doing sharpening first H. This is what you see processing astronomical images - the PI guys argue any deconv should always be done first with subtle settings, followed by curves, saturation, sharpening etc as required. You also need to have low noise in the image to start with (=stacks of data in Astrophotography) or else you just clump the background noise.
Darrin's examples interesting too - certaintly how much the fence wire comes out afterwards.
25-04-2012, 05:22 PM
If the image is sharpened first, then, the deconvolution (or whatever algorithm Canon is employing) is only going to exacerbate the sharpened image and induce halos/artefacts. :)
I've tested this myself and it seems to make sense -- it's subtle, but, the artefacts appear.
Subtle sharpening post-DLO is fine.
vBulletin® v3.8.7, Copyright ©2000-2013, vBulletin Solutions, Inc.