LX3 Optimal RAW Processing: High Resolution

Started May 23, 2013 | Discussions thread
OP fPrime Senior Member • Posts: 2,258
Re: LX3 Optimal RAW Processing: High Resolution

Detail Man wrote:

Hi there, fPrime,

I get no thrills if and when others may use DxO Optics Pro (as I rather like being different) - but since we each processed your LX3 RW2 rather differently, I don't really think it is highly informative as a comparison due to differences in Brightness and Contrast. As I wrote to you in March:

It's just like I said it would be. Our processing preferences and tastes differ to the extent that the images are not comparable. At different maximum luminance and different contrast levels (different RGB tone-curve transfer functions), color saturation levels, etc., each processor might produce results more similar to the other.

DM, thanks again for helping with the DxO conversion.  I realize you think that overall the image is not comparable because your aesthetic choices are different from mine, but I was hoping you could comment specifically on the other RAW processing differences I noted.

For example, it would be interesting to hear your thoughts on DxO's lack of fine detail color preservation in the parked cars and with the yellow flowers.  I suspect these were not caused by different RGB tone transfer curves or darker luminance choices.  Some of the colors were crushed so much that I doubt even vibrancy or saturation boosts would recover them.

Another thought is that if Jimmy has used SP 5 "default" settings, that may not be representative.

Note the significant difference between the DxO and ACR histograms, and the overall appearance:

DxO Optics Pro 7.23

Adobe Camera RAW 6.7

The DxO processing is linear (within the range of RGB). The ACR processing clips all RGB channels.

Agreed, my ACR rendition may well clip the RGB channels (presumably in the highlights of the sky) but I can't see any practical loss of image quality from that.  In fact, I like that fact that the brighter rendition allows me to see deeper into the shadow areas of the image where real detail matters to the human eye.  To me the ACR conversion appears more natural.  I'm sure you already noticed that your linear DxO rendition is noticeably darker than those from Silkypix, UFRaw, and ACR.  I'm not saying that linear RGB processing is wrong, I'm just saying it doesn't look as good on this image. 

When I linearize the numerical ratio of the average gray-levels (105/88) of the above two histograms in order to remove the non-linear effects of the gamma-correction applied by mapping into sRGB color-space, I get a 0.534 EV ("stop") difference in overall image brightness. Together with the existence of the steeper tone-curves used in the DxO processing (relative to the PS with ACR processing), those represent significant numerical and visual differences.

You probably wished that I had processed your image to have an overall image brightness (and perhaps contrast of the RGB tone-curves, as well) closer to your choices for the purposes of such a comparison. I'm not able to do other than follow my own aesthetic view, however ...

True, but aesthetics aside we can still study fine color detail preservation, demosaicing artifacts, sharpening artifacts, lens correction, etc.

While the DxO Optics Pro 7.23 processing involved only 16-bit TIFF export, application of very mild small radius, low strength USM in PSP X4 SP3, then loss-less JPEG encoding using XnView, you mentioned that your processing included the following applications: Photoshop (with ACR 6.7 plug-in), Photo Wiz's ColorWasher, Focus Magic, and PTLens. Probably worthwhile to mention.

You're stealing some of my thunder here, but yes I do still need to outline my ACR process.  That's coming up next.

I also noted in March:

The demosaicing in LR 3.x / ACR 7.x and LR 4.x / ACR 7.x is (in the case of some cameras, it may well potentially be different for any given camera) may be a bit better for the tiniest of details ("nano-detail") - whereas DxO can be better for "micro-detail". I think that part of that may well be that DxO implements the deconvolution-deblurring component of it's "Lens Softness" corrections as well as some aspects of it's Luminance NR at the RAW level (integrated as part of the de-mosaicing processes).

The cropped samples that you are extracting appear to be (around) 250x250 pixel sized samples. That allows (at very best) resolution of only around 83 line-pairs per height/width. Not much for spatial frequency resolution, given the level of details that you hope to present for inspection ? Up-sampling the crops by a factor of 2 is not adding any new information (save for possible re-sampling artifacts if you are not using the Nearest Neighbor algorithm to do so).


In the age of DxO 8.x and LR 4.x (ACR 7.x), both of these newer applications have automatically calculated localized tone-curve transfer-functions that may well offer better results. I have seen LR 4.x achieve results in those specific respects that DxO 7.x definitely cannot.

If LR 4.x ran on my WinXP, I might be tempted. Not going to be installing CS6 just to run ACR 7.x, though - especially in light of Adobe's latest shenanigans. DxO 8.x is a 64-bit item that they claim will run efficiently on my (32-bit) Win XP Pro - but I am skeptical of their claims, DxO 8.x has some changes that I don't think that I would like, and they (DxO Labs) are jerks, too.

I'm with you on that.  Given Adobe's predatory business model I don't think I will be upgrading to PS6 CC so obviously will be orphaned at ACR v6.7 for some time.


So, with DxO Optics Pro 7.23 supporting my LX3 and GH2 with both my lenses, I'll stick with it.

DM ...

 fPrime's gear list:fPrime's gear list
Canon EOS 5D Fujifilm FinePix S5 Pro Nikon D1X Nikon D200 Nikon D700
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow