fredmiranda.com Login
This topic is in the 2007 archive. You can read it and like its posts here. It is read only.

Archive 2007Canon DPP v3.2 IQ vs. v3.0.1

  
 
rscheffler
Online

p.1 #1Canon DPP v3.2 IQ vs. v3.0.1
I just posted this in another thread, but thought it might get lost in that discussion since it's way at the end. https://www.fredmiranda.com/forum/topic/585219/5/#5131732

Maybe this is more appropriate to the Post-processing & Printing forum, but this one probably gets a lot more traffic, so I'll try here first.

Hopefully this is acceptable with the moderators.

Anyway, I noticed some differences in v3.2 conversions (with 1D Mark III files) compared to v3.0.1.

------------------

I have some initial reservations about certain aspects of v3.2.

I just used v3.2 to re-process a few images taken with the Mark III and 16-35II to fix some minor CA and noticed that the corrected images are significantly softer in certain areas. Specifically, the v3.2 correction seems to degrade sharpness/detail in certain colors. Here are some samples - look particularly at the detail in the yellow areas:

The entire image:

http://www.pbase.com/scheffler/image/88434604.jpg

DPP 3.0.1 conversion near the frame edge:

http://www.pbase.com/scheffler/image/88424155.jpg

DPP 3.2 from the same spot with CA and some vignetting correction:

http://www.pbase.com/scheffler/image/88424158.jpg


From the middle of the image:

DPP 3.0.1:

http://www.pbase.com/scheffler/image/88424161.jpg

DPP 3.2:

http://www.pbase.com/scheffler/image/88424162.jpg


These conversions have sharpness set to zero in DPP. Applying USM in Photoshop will bring back some detail in the orange and yellow areas of the 3.2 conversions, but never to the point of the sharper 3.0.1 conversion (which obviously sharpens up much better with USM than the 3.2 version).

BUT... looking at this some more, I don't feel it's the CA correction that is causing the change in image quality

I looked at images from v3.0.1 and v3.2 without CA correction in the red and blue channels and noticed something interesting. In v3.0.1 conversions, the red channel is not very sharp, but in v3.2 conversions, detail is very good. In the blue channel it's the opposite. v3.0.1 is very sharp but the v3.2 blue channel is somewhat smeared. Based on a few preliminary image comparisons, it appears that v3.0.1 and 3.2 each handle the red and blue channels differently (and also the green channel). In v3.0.1 the red channel is softer than v3.2 but the green is slightly sharper and the blue is significantly sharper. When comparing the same image converted in each version, overall the 3.2 version is ever so slightly sharper except in areas with a lot of yellow and orange (where the blue channel is quite dense). Overall the v3.2 conversions I've looked at so far (just a handful) seem to be fine and appear to be slightly crisper thanks to the sharper red channel, which also helps control red "bleed" into neighboring areas when there is a sharp transition... though the DPP jaggies are still there with sudden transitions to red areas.

So, that made me wonder how an image would look if I pasted the sharper v3.2 red channel into the v3.0.1 image's red channel.... and it works, resulting in an image that is definitely better than the straight v3.0.1 image and slightly sharper than v3.2. And if CA has been applied to the v3.2 conversion, that also carries through somewhat in the red channel.

So far I've only made a few conversions, but this immediately caught my eye due to the abundance of orange and yellow in some of the images.

Reading through the 3.2 help manual, I also noticed the statement about distortion correction: "After correcting Distortion, the image may appear with less resolution. If this occurs, you can make the image similar to the one before correction by applying more sharpness with sharpness adjusting function..." which leads me to wonder, just how much sharpness is lost?

So I did a comparison and it looks like the loss in sharpness is there but not a lot. Each of us will have to decide whether it's acceptable. For most applications I think it is acceptable, based on initial impressions looking at a few images. Additionally, images corrected for distortion are output at a larger dimension than usual. For the Mark III it's 3978x2652 compared to 3888x2592.

Ron

BTW, here are some screen grabs at 300% of red and blue channels:


Note the detail difference in the red strap that runs vertically through this crop
v3.0.1 red:

http://www.pbase.com/scheffler/image/88429865.jpg

v3.2 red:

http://www.pbase.com/scheffler/image/88429867.jpg

This is a crop from the orange and yellow vest in the middle of the image
v3.0.1 blue:

http://www.pbase.com/scheffler/image/88429869.jpg

v3.2 blue:

http://www.pbase.com/scheffler/image/88429871.jpg

Nov 04, 2007 at 05:47 PM
Dave In La
Offline
Details
p.1 #2Canon DPP v3.2 IQ vs. v3.0.1
interesting.....I just started using v3.2.

definitely using same color for process?
Nov 04, 2007 at 06:13 PM
rscheffler
Online
Details
p.1 #3Canon DPP v3.2 IQ vs. v3.0.1
The settings were identical. I even went back to v3.0.1 to process some different files and then processed those through v3.2 with identical settings. I saw the same differences.

Overall I don't think it's going to significantly affect most images. I can foresee some benefits to a smoother blue channel, like sunsets where you don't really need much detail sharpness.

Ron
Nov 04, 2007 at 06:58 PM
esanchez
Online
Details
p.1 #4Canon DPP v3.2 IQ vs. v3.0.1
The only thing I have noticed is the now picture style faithful seems more accurate then Neutral. Before I used to set my camera to Standard and convert/change to Neutral because it would give me the best colors. Now Neutral seems kind of flat and dull while Faithful gives me the accurate colors especially skin tones. I'll have to look into the color detail you mention....
Nov 04, 2007 at 07:03 PM
Tom_W
Offline
Details
p.1 #5Canon DPP v3.2 IQ vs. v3.0.1
Interesting - I'll be watching this thread.
Nov 04, 2007 at 07:56 PM
tmr4
Offline
Details
p.1 #6Canon DPP v3.2 IQ vs. v3.0.1
I noticed this also. The help manual hints at this as follows:

"After correcting [Peripheral illumination], some noise may appear in the periphery of an image. If this occurs, you can reduce the noise by using luminance and/or chromatic noise reduction of the noise reduction function.

After correcting [Distortion], the image may appear with less resolution. If this occurs, you can make the image similar to the one before correction by applying more sharpness with sharpness adjusting function.

After correcting [Color blur], the colors of the image may become weaker. If this occurs, you can make the image similar to the one before correction by making the color deeper with adjusting color tone and/or color saturation functions.

[Color blur] correction is effective for correcting blue or red color blur which sometimes occurs at the edge of the highlighted area of an image. If conditions do not match, correction cannot be made. "
Nov 04, 2007 at 08:08 PM
rscheffler
Online
Details
p.1 #7Canon DPP v3.2 IQ vs. v3.0.1
esanchez wrote:
The only thing I have noticed is the now picture style faithful seems more accurate then Neutral. Before I used to set my camera to Standard and convert/change to Neutral because it would give me the best colors. Now Neutral seems kind of flat and dull while Faithful gives me the accurate colors especially skin tones. I'll have to look into the color detail you mention....


I'll keep an eye on this since I use mostly Faithful and Neutral but in the few conversions I've done so far, and only with 1DIII files, I haven't noticed any difference. The comparisons I've done have been by layering different conversions in one Photoshop file and toggling the layers on and off. In this manner it's very easy to pick out subtle changes.

If you're seeing differences but I'm not, it could well be camera model dependent In past updates using older cameras like the 1DII and 1DIIN I did notice slight changes in image rendering.

Ron
Nov 04, 2007 at 09:03 PM
Dave Jr
Offline
Details
p.1 #8Canon DPP v3.2 IQ vs. v3.0.1
Part of what I see in your 100% crops is that the 3.2 conversions seem to have slightly higher exposure in the highlights (or slightly more contrast), which is blowing the detail. Look at the guy's face, and sweatshirt, and the white stripes in his oxford, they are all brighter in the 3.2 version. Can you get them to look more similar (i.e. bring back the highlight detail) by either notching down the exposure or contrast slighltly?

Nov 05, 2007 at 09:22 AM
Mel Gross
Offline
Details
p.1 #9Canon DPP v3.2 IQ vs. v3.0.1
It's possible that among other things, Canon is conceding that the blue channel is noisier than the others, as is normally the case. Therefore, they are emphasizing sharpness in the red channel, to the exclusion of the blue, in an attempt to keep that noise down.

Most recommendations when sharpening, and for noise reduction, is to blur the blue, and sharpen the channel that is the least noisy the most, followed by the second least noisy.

To tell the truth, I find Canon's software to be too complex and not as useful as Adobe's Camera Raw, particularly version 4.2.

As I've seen in comparison tests, Canon and Nikon, with their own software, are much less likely to attempt to wring that extra quality out of their files, if it means that they have to do anything that smacks of estimation, such as effective hi-lite recovery.

This could be another example of this, as it looks as though Canon is taking a very conservative approach to both sharpening, and noise reduction.
Nov 05, 2007 at 09:40 AM
ns66
Offline
Details
p.1 #10Canon DPP v3.2 IQ vs. v3.0.1
interesting
now the latest version is 3.3.1.1, i wonder is this 'fixed'
May 31, 2008 at 01:51 PM
James Burden
Offline
Details
p.1 #11Canon DPP v3.2 IQ vs. v3.0.1
3.4.1 is the latest for OS X at least....
May 31, 2008 at 02:57 PM
EB-1
Offline
Details
p.1 #12Canon DPP v3.2 IQ vs. v3.0.1
Weird. I don't have a 1D MK III, but the DPP 3.3 output from 1Ds MK III files do not show any disparity in the blue channel detail. Did you update to v3.3.1.1 yet? It has been out for a while now.

EB
May 31, 2008 at 03:16 PM
rscheffler
Online
Details
p.1 #13Canon DPP v3.2 IQ vs. v3.0.1
I've since updated to 3.4.1 and haven't looked at it so closely. In the end it really hasn't matter that much to me since each successive version has added features that have generally outweigh any slight changes in RAW file conversion.
I'm out of town for a while, but maybe I'll dig up that file once I'm back and convert it with 3.4.1 to see if there are any new changes...

Ron
May 31, 2008 at 06:33 PM
thepiecesfit
Offline
Details
p.1 #14Canon DPP v3.2 IQ vs. v3.0.1
ive noticed some odd jaggy pattern in most new versions of dpp. Ive started a thread about it here. i do hope canon fixes this eventually http://forums.dpreview.com/forums/readflat.asp?forum=1032&thread=28683103
Jul 21, 2008 at 05:20 PM
akclimber
Offline
Details
p.1 #15Canon DPP v3.2 IQ vs. v3.0.1
Interesting observations. Thanks!
Jul 21, 2008 at 09:39 PM
Gary Petersen
Offline
Details
p.1 #16Canon DPP v3.2 IQ vs. v3.0.1
Version 3.4.1.1 (windows) is what came with my 1D.
Jul 21, 2008 at 09:52 PM
FM ForumsCanon ForumJoin Upload & Sell
  
 

You are not logged inLog inRegister