requesting dxo raw workflow tips

gteague said:

thanks for the link. even translated from russian it looks like some very useful information. i will file it away and revisit it if i ever build up enough background in digital imaging to actually make use of it.

guy

exdeejjjaaaa said:

for dxo lovers = http://momentextractor.livejournal.com/43968.html

if you are too lazy to translate then just watch animated illustrations

Click to expand...

Click to expand...

Detail Man said:

Have Googled the subject of ImageMagick Lanczos resizing, and it appears to be a complicated subject in itself. I may (or may not) figure this out. I have not used ImageMagick - because I am lazy about command-line stuff in my old age, and am spoiled by GUIs. You would want to make sure that it is Lanczos-3, and I am not sure that the "distort" switch is a good idea to use. Dunno. I will read a bit about this - but the ImageMagick documentation and Forum are always confusingly deep for me with my limited familarity with it. Here's the web-page for IM "switches":

http://www.imagemagick.org/script/command-line-options.php

Would suggest that you try using DxO's Bicubic to make 16-bit TIFs, and see if things get better. I know that XnView and FastStone use Lanczos-3. They are 8-bit, but that may work out for you.

I use an old version of Artizen HDR 2.86 that features 16-bit Lanczos-3. It's a weird HDR application that also has individual 16-bit editing capabilities. I just use their resizing tool only. Their newer 3.x version (renamed "Photoroom", I think) only does 8-bit processing in the trial version. They have a "Lanczos Sharper" option that was causing undesirable artifacting for me - so now I only use their standard Lanczos (which I managed to determine from their support is Lanczos-3).

In the end, the DxO 16-bit Bicubic might serve you well enough for your tastes (or the 8-bit Lanczos-3 in XnView or FastStone ). I tend to be pernsnickety, and into esoteric and (perhaps) sometimes "masochistic" rigors. Have a feeling that the ImageMagick resing may be causing you extra problems with artifacts. I would not use it unless you fully understand the command-line "switches", and how to configure it for Lanczos-3, as well as to use an appropriate "window" (which is the anti-aliasing low-pass filter that I previously described precedes re-sampling algorithms)

Click to expand...

Detail Man said:

Detail Man said:

Have Googled the subject of ImageMagick Lanczos resizing, and it appears to be a complicated subject in itself. I may (or may not) figure this out. I have not used ImageMagick - because I am lazy about command-line stuff in my old age, and am spoiled by GUIs. You would want to make sure that it is Lanczos-3, and I am not sure that the "distort" switch is a good idea to use. Dunno. I will read a bit about this - but the ImageMagick documentation and Forum are always confusingly deep for me with my limited familarity with it. Here's the web-page for IM "switches":

http://www.imagemagick.org/script/command-line-options.php

Would suggest that you try using DxO's Bicubic to make 16-bit TIFs, and see if things get better. I know that XnView and FastStone use Lanczos-3. They are 8-bit, but that may work out for you.

I use an old version of Artizen HDR 2.86 that features 16-bit Lanczos-3. It's a weird HDR application that also has individual 16-bit editing capabilities. I just use their resizing tool only. Their newer 3.x version (renamed "Photoroom", I think) only does 8-bit processing in the trial version. They have a "Lanczos Sharper" option that was causing undesirable artifacting for me - so now I only use their standard Lanczos (which I managed to determine from their support is Lanczos-3).

In the end, the DxO 16-bit Bicubic might serve you well enough for your tastes (or the 8-bit Lanczos-3 in XnView or FastStone ). I tend to be pernsnickety, and into esoteric and (perhaps) sometimes "masochistic" rigors. Have a feeling that the ImageMagick resing may be causing you extra problems with artifacts. I would not use it unless you fully understand the command-line "switches", and how to configure it for Lanczos-3, as well as to use an appropriate "window" (which is the anti-aliasing low-pass filter that I previously described precedes re-sampling algorithms)

Click to expand...

NikonGuyUK ,

Here is an interesting web-page that I thought of when you brought up "moire-like" artifacting:

http://bvdwolf.home.xs4all.nl/main/foto/down_sample/down_sample.htm

Unfortunately, it does not go into ImageMagick re-sampling command details - and all the ImageMagick Forum stuff that I can find is rather esoteric, flies over my head, and manages rather well to fail to clearly spell-out the command syntax for dummies every time. There must be something out there about ImageMagick that I can understand, but I never seem to be able to find it whenever I have searched. I would be interested if you find anything tangible about it !

Click to expand...

henrywho said:

henrywho wrote:

1) Downsampling is better done in the RGB colorspace but not sRGB colorspace. And I believe the same for USM.

Click to expand...

henrywho said:

2) Blur=0.88549061701764 for Lanczos3 is probably too sharp and prone to moire, especially when followed by USM.

Assuming that your input in sRGB, try:

convert %1 -colorspace RGB -define filter:filter=Jinc -define filter:window=Quadratic -define filter:lobes=3 -distort resize 1200x1200 +repage -unsharp 0x0.75+0.75+0.008 -colorspace sRGB %~n1_JQ3_USM.tif

(and tune the USM strength as necessary)

And, though not widely acknowledged, I like EWA Lagrange more than USM in many cases (except for really tough cases like "backpack"):

convert %1 -colorspace RGB -filter Lagrange -distort resize 1200x1200 +repage -colorspace sRGB %~n1_Lag.tif

All credits go to NicolasRobidoux in the ImageMagick forum.

Click to expand...

henrywho said:

henrywho wrote:

Talking about RAW Therapee, do you know if there is any lens profile for Oly 45mm f1.8, Pan 20mm f1.7 and Oly 12mm f2.0?

Click to expand...

Detail Man said:

Detail Man wrote:

Detail Man said:

gteague wrote:

i have purchased dxo software and an update released yesterday has a module for the em5. they already had the gx1 and most of the lenses i own. i decided on dxo because with a mix of oly and panasonic bodies and lenses, dxo seems to be the solution to correct for ca and distortion no matter what lens is on what body.

but i also have iphoto and aperture and graphic converter and ps elements 9 and have the trial version of lightroom installed. what i need to know is whether i can use dxo for everything from raw to finished image or whether i'd be better off introducing lightroom or aperture or the other software i listed into the workflow. a brief outline or flow chart or a pro/cons table would be extremely helpful. once i have an idea of what product(s) i need to use, i can do the rest of the research to learn how to use the products i need.

Click to expand...

DxO Optics Pro has good quality de-mosaicing (in some cases perhaps not as good as Lightroom, but probably pretty closely comparable overall). Both programs have good quality de-mosaicing.

The high-quality automatic/manual Rectilinear Distortion corrections and the unique and (I find) quite useful automatic/manual "Lens Softness" corrections are (assuming that the image is not overly de-focused in the first place) very useful. "Lens Softness" does not correct for poorly focused shots, however. It corrects for lens-system weaknesses and certain aberrations very well.

See: http://www.dxo.com/us/photo/dxo_optics_pro/features/optics_geometry_corrections/lens_softness


I much prefer using DxO "Lens Softness" to the Sharpening tools in Lightroom 3.x (which has the very same Sharpening and NR tools as does Lightroom 4.x). For more highly de-focused images, it appears that Lightroom 3.x/4.x Sharpening tools can (what I call "fake") image-detail better than DxO. (I think) that this is a result of deconvolution-deblurring artifacts - artifacts that (I think) can really look ugly fast (whenever the "Detail" control-slider is set to greater than around 25 or so).

The automatic/manual Chromatic Aberration and Purple Fringing and Vignetting corrections are pretty good - Lightroom 4.x (might) accomplish the CA and PF stuff better (not sure about that).

My OS is WinXP pro - so I am not able to install and use Lightroom 4.x.

DxO Optics Pro lacks meaningful "highlight recovery" provisions. There is a Tone Curve tool, but only small and gradual "shoulders" at the very top of the tone-curve seem to work for me. Lightroom 4.x's abilities in that regard do indeed appear to be much more effective (particularly so if you like to "blow" highlights). It's unclear how well it works on finely-detailed subject-matter ...

The DxO Lighting tools are something that I find to be often quite useful for tone-curve adjustments. I really like the ability to adjust the Gamma (something that Lightoom 3.x/4.x does not feature). On the other hand, it appears that Lightroom 4.x does things with tone-curve modifications that (may well) be performed on a "local" (not just "global") basis. Lightroom 4.x also has some "selective processing" abilities that DxO lacks. This is an important difference for some ...

DxO Optics Pro has pretty good NR. When the Luminance NR is set to around 10 or less, it appears to preserve more fine-detail than does Lightroom 3.x/4.x Luminance NR. For higher image-noise tasks, I sometimes export a 16-bit TIF from DxO, and use Lightroom 3.x/4.x's excellent Color NR to do the NR. The Lightroom 3.x/4.x Luminance NR is just OK - about the same as anybody else's.

I would think that Lightroom 4.x (which can be purchased for $150 USD or less when on sale) would be a worthy additional application for you to own. You purchsed DxO for only $99 USD, so (perhaps) you have some cash left over.

DxO is well worth using as a front-end (for it's optical corrections, including "Lens Softness"). You may want to export that initial processing as 16-bit TIF to Lightroom 4.x in order to take advantage of it's probably (overall) better tone-curve adjustment capabilities (including much-praised) "highlight recovery". The quality of (particularly) the Color NR is somewhat better than DxO (esp. for higher image-noise cases). Using DxO Optics Pro in these ways is not unusual.

Here is a thread that I created for a friend a while back (about DxO 6.x, but still very applicable to DxO 7.x) that may provide somewhat interesting and somewhat informative reading for you:

http://forums.dpreview.com/forums/readflat.asp?forum=1033&message=38343344


regards,

DM ...

Click to expand...

Detail Man said:

Detail Man wrote:

Detail Man said:

gteague wrote:

so, one dumb question. everything i've ever read advises to do sharpening at the very end of the post processing. i'm pretty sure i already know the answer to this since i read some of the other thread you cited, but does performing this dxo process of sharpening at the start of the workflow (assuming i start with dxo and export to lr4 via tiff) apply to that rule?

Click to expand...

An excellent question, really. The usual regimen that makes sense is to do NR early-on in the processing-chain, and (indeed) to perform Sharpening last. This remains true as a general rule.

The rules are a bit different where it comes to deconvolution-deblurring processes (as are integrated into "Lens Softness" corrections. It is more of an initial correction for lens-system weaknesses than it is more along the lines of the "classic" subtraction of a blur-mask from the original (as in Unsharp Masking and High-Pass Sharpening).

Not sure, nut it may also be that some of the elements of "Lens Softness" are integrated into the initial (necessarly first in the chain) de-mosaicing process. This is also (likely) the case where comes to certain elements of the Luminance Noise reduction processes (where fine-detail information may be extracted from de-mosaicing, and re-added in the application of Luminance NR - in order to help to preserve fine-details).

Whereas in Lightroom there exists a large number of interacting NR and Sharpening control-sliders that have to be tweaked in a complicated dance of control-tweaking, the process is much more straighforward in DxO Optics Pro. It is this simple. Get the "Lens Softness" controls where you want them as early on as you like in the processing-chain. Later, after all else is tweaked, have a close look (at 200% or at 400%), and adjust the NR controls to remove image-noise as needed.

There really is very little interaction between the two tools. (Sometimes), I may iterate a bit (after playing with th NR controls late in the process). Also, when image-noise is at higher levels (and or the ISO Gain of the image is higher), DxO automatically adjusts the Lens Softness parameters somewhat (remember, it is a very "smart" program, just like they claim). I usually tend to like the adjust the settings of the NR controls a tad higher to smooth out any deconvolution-deblurring artifacts arising out of the "lens Softness" corrections. I prefer that to increasing the (newly added) "Bokeh" control (more than around 25 out of 100, anyway. The "Bokeh" control performs some amount of spatial-frequency low-pass filtering, and I find that sparing anount of the Chrominance and Luminance NR can do a better job (without sacrificing fine-details as much).

A note about the DxO NR controls. I have mostly used it with LX3 RW2s and GH2 RW2s. In both cases, at least in the "maximized exposure - minimized ISO Gain" shooting that I do, I am able to set the Chrominance and Luminance NR controls to mere fraction of the settings that the program automatically sets. See the "DxO for Doto" thread posts for a table of typical values of NR settings.

I often (but not always) after down-sampling my 16-bit TIF output image files from DxO before converting them to JPGs will apply very mild (16-bit) USM to my images. USM Radius=0.5 ranging only to 0.7 or a maximum of 1.0 in very soft images. USM Strength=50%, ranging up to 100% when feasible, or down to 25% for highly detailed subject-matter. USM Threshold=between 3 and 6 (out of a maximum of 100). All down-sampling algorithms are pre-ceded by digital anti-aliasing filters to attenuate moire'-like effects in the final down-sampled output. Mild USM restores the small reductions in image-sharpness that result frm that. Also, if your JPEG encoding to follow uses less than 100% Quality Factor, the mild USM helps to "goose" the sharpness (slightly) to help to pre-compensate for those effects.

My typical approach is to almost never use the DxO USM tool at full size. I use the "Lens Softness" (typicall around "Global"=0, and "Detail"=0), not pushing things too much. After exporting a 16-bit TIF from DxO, and usning a 16-bit Lanczos-3 down-sampling algorithm that I like to use, I use PSP X4 to perform mild (16-bit) UM. Then I use XnView's excellent JPEG encoder to make a final JPG.

Lots of good information offered (I hope). Now everybody knows my secrets - and they will (no doubt) ignore them completely, and do things "their way" ... ... Fine by me. I like what I get.

Click to expand...

Detail Man said:

Detail Man wrote:

Some users rightfully note that DxO Optics Pro has a tendency to (by default) impart too much contrast to images - in particular, causing shadow-areas to appear as too dark for their liking.

There are various ways to reduce this behavior:

(1) (DxO Lighting), reduce the Black Point setting from what is automatically selected;

(2) (DxO Lighting), reduce the Preserve Shadows control-setting below it's default value of 100;

(3) (Color Rendering), reduce the setting of the Intensity control below 100 in order to reduce the slope (and thus the contrast) of the factory in-camera JPG tone-curve transfer-functions;

(4) (Vignetting) avoid setting the Shadow/Highlight Preservation near maximum (the control has very little effect on Highlights, but can and does have significant effects upon Shadows);

(5) (DxO Lighting), be sparing on the level of the Global Contrast control (settings of 10 or less).

By the way, the Local Contrast function (in DxO Lighting, and which is always set to a setting-value of 20) is a USM-related process that I seldom find useful except in cases of "emergency rescue operations" on images that are very low-contrast and or foggy/hazy. I always disable it (setting it to Zero), as it can decrease the quality of the "Lens Softness" image acuity enhancement.

Another note is that when DxO Lighting selects a White Point value that is less than it's maximum setting-value of 155, the R, G, and B channel histograms are "stretched", which is bound to result in greater processing artifacts than simply using the Exposure Compensation and the Gamma control in order to set overall image-brightness. Thus, I normally set it to 255.

Note that the indicated WB Color Temperature reference reading in DxO Optics Pro seems from my experience to correspond to a setting of what is around 500 Degrees K higher in Adobe LR/CR.

Some users rightfully note that DxO Optics Pro has a tendency to (by default) impart too much emphasis on Green hues (sometimes causing foliage and grass to appear unrealistic and gauche). When I see this in my in-process image, I use a standard H/S/L Tool Saturation curve that boosts Blue and cuts Yellow, using linearly-related slope values for the four other colors that form a linear "triangular" function in numerical value of the Saturation settings of those colors.

Click to expand...

Detail Man said:

Detail Man wrote:

Detail Man said:

Detail Man wrote:

Some users rightfully note that DxO Optics Pro has a tendency to (by default) impart too much contrast to images - in particular, causing shadow-areas to appear as too dark for their liking.

There are various ways to reduce this behavior:

(1) (DxO Lighting), reduce the Black Point setting from what is automatically selected;

(2) (DxO Lighting), reduce the Preserve Shadows control-setting below it's default value of 100;

(3) (Color Rendering), reduce the setting of the Intensity control below 100 in order to reduce the slope (and thus the contrast) of the factory in-camera JPG tone-curve transfer-functions;

(4) (Vignetting) avoid setting the Shadow/Highlight Preservation near maximum (the control has very little effect on Highlights, but can and does have significant effects upon Shadows);

(5) (DxO Lighting), be sparing on the level of the Global Contrast control (settings of 10 or less).

By the way, the Local Contrast function (in DxO Lighting, and which is always set to a setting-value of 20) is a USM-related process that I seldom find useful except in cases of "emergency rescue operations" on images that are very low-contrast and or foggy/hazy. I always disable it (setting it to Zero), as it can decrease the quality of the "Lens Softness" image acuity enhancement.

Another note is that when DxO Lighting selects a White Point value that is less than it's maximum setting-value of 155, the R, G, and B channel histograms are "stretched", which is bound to result in greater processing artifacts than simply using the Exposure Compensation and the Gamma control in order to set overall image-brightness. Thus, I normally set it to 255.

Note that the indicated WB Color Temperature reference reading in DxO Optics Pro seems from my experience to correspond to a setting of what is around 500 Degrees K higher in Adobe LR/CR.

Some users rightfully note that DxO Optics Pro has a tendency to (by default) impart too much emphasis on Green hues (sometimes causing foliage and grass to appear unrealistic and gauche). When I see this in my in-process image, I use a standard H/S/L Tool Saturation curve that boosts Blue and cuts Yellow, using linearly-related slope values for the four other colors that form a linear "triangular" function in numerical value of the Saturation settings of those colors.

Click to expand...

Have been reading the User Guide for DxO Optics Pro Versions 8.x, and can see that some (but not all) of the above text refers to the versions 7.x "DxO Lighting" tools ("Black Point", "White Point", and "Preserve Shadows" above). That ("DxO Lighting") interface is available in DxO 8.x as an option.

"Local Contrast" has been renamed "Microcontrast". "Global Contrast" has been named "Contrast".

The "Smart Lighting" and the "Smart" button in the "Exposure Compensation" tool don't appeal to me - be neither does any camera or processor/editor's claim that it can think for me.

I don't know whether the Selective Tone controls are something useful (or how they compare to the very similar looking controls in LR 4.x / CR 7.x). I read one experienced user state in a post that the tendency for prior versions to overdo the shadows has been remedied by the ability to adjust the "Black" control down to a zero black-point. That could be useful.

Other than that, I get the feeling that I might (myself) possibly miss the Version 7.x "DxO Lighting" controls. It's a good thing that those Versions 7.x user-controls remain an option in Versions 8.x.

Click to expand...

Detail Man said:

Detail Man wrote:

Detail Man said:

henrywho wrote:

1) Downsampling is better done in the RGB colorspace but not sRGB colorspace. And I believe the same for USM.

Click to expand...

It makes sense to me that both re-sampling and edge-enhancement operations would be best accomplished in linear units (as opposed to proceeding in non-linear gamma-corrected units).

Detail Man said:

2) Blur=0.88549061701764 for Lanczos3 is probably too sharp and prone to moire, especially when followed by USM.

Assuming that your input in sRGB, try:

convert %1 -colorspace RGB -define filter:filter=Jinc -define filter:window=Quadratic -define filter:lobes=3 -distort resize 1200x1200 +repage -unsharp 0x0.75+0.75+0.008 -colorspace sRGB %~n1_JQ3_USM.tif

(and tune the USM strength as necessary)

And, though not widely acknowledged, I like EWA Lagrange more than USM in many cases (except for really tough cases like "backpack"):

convert %1 -colorspace RGB -filter Lagrange -distort resize 1200x1200 +repage -colorspace sRGB %~n1_Lag.tif

All credits go to NicolasRobidoux in the ImageMagick forum.

Click to expand...

Thanks for the input ! Not being a user of ImageMagick, my visits to their forum (after seeing what I can see in the published user-guides) has been a confusing (though not highly unusual in such cases of users discussing intricate nuts and bolts of similar applications) experience.

My visits to the RAW Therapee forum are similar. Without being deep into the innards as a user of an ever-developing application, trying to follow all of the factoids, patches, and modifications feels a bit like grasping at straws. One may see this or that which (may) be relevant - but it is less than clear what is current, and thus cogent. It requires more diligence than I usually have.

Click to expand...

Nicolas Robidoux said:

Nicolas Robidoux wrote:

Template for an ImageMagick GUI

Click to expand...

Detail Man said:

Detail Man wrote:

Click to expand...

Detail Man said:

Are you familiar with the (recently created) Photographic Science and Technology forum (?):

http://www.dpreview.com/forums/1061

Click to expand...