Jack, originally I used Camera Raw CS2, and 16-bit ProPhoto RGB. Later I moved to Lighroom, where I can only use 16-bit Adobe RGB.
Pondria, your claim that the processing is largely in the Raw converter is only true for the theoretical case of infinite precision in the raw files. When we are dealing with limited bit-depths and finite spaces, the accuracy of the recorded values, as well as proper calibration, becomes much more important than which raw processor we use.
steven, the DMR thread got deleted (there is some work under way to save all the info from various caches and places, and put it somewhere off-site, as well as the possibility that moderators may pull it back into a normal archived thread). The "M8 (only) Bible" thread was combined with the "Hands-On" thread by moderators who did not want two similar threads running.
Back to the topic of colour and RAW etc - for my sins I've been processing about 1000 images (not mine) tonight taken on a 20D and NikonD200 - most were RAW but some on the 20D were JPEG and I'm gobsmacked by how the skin tone is looking straight out of the camera with what the Exif data tells me are 'standard' parameters. For some reason the photographer jumped between RAW and Jpeg on the same scene and try as I might with both C1 and ACR I can't match the quality of the skintone in the jpeg with either of them. Me thinks Canon knows what it's doing with it's in camera processing and ain't letting on to Adobe or PhaseOne etc. I need to look into this further but this has upset the applecart for me tonight at least.
carstenw wrote:
Jack, originally I used Camera Raw CS2, and 16-bit ProPhoto RGB. Later I moved to Lighroom, where I can only use 16-bit Adobe RGB.
Just in case you aren't aware, the Canon 5D's gamut is actually larger than Adobe RGB, especially in the high yellows to reds. The biggest range however is the yellows and this can of course affect how reds render, depending on how the colorspace conversion engine converts. If it uses perceptual rendering intent, the colors outside red will cause an increase in saturation in the visible reds. Relaitve colorimetric is significantly better, but still can show some over saturation in very bright reds. The best bet is to convert out to ProPhoto from RAW and do your colorspace conversions in PhotoShop using RC with BPC. However, if you feel you must continue to use Lightroom/ARGB, then the best you can do is desaturate the red channel as you are doing. Actually, I haven't tried it myself, but I suspect if you desaturate yellow by about twice what you desaturate the reds by (and get to about the same total desaturation amount you currently use for red) you will probably find a happier result than just by doing the red channel alone .
Cheers,
Edited by Jack Flesher on Dec 20, 2006 at 04:12 PM GMT
Jack, I was not aware that the 5D has a larger gamut than Adobe RGB, but why then does Canon only support Adobe RGB in the camera, and not ProPhoto RGB? That sucks. Now that I think about it, the worst shot I had (the red tulip shot), I developed with Lightroom.
That really sucks, because I like Lightroom *so* much better than Camera Raw, and I am not about to drop €500 on C1. I like Raw Developer, but it has little work-flow to compare with Lighroom. Aperture is not ready yet, although it sounds like it is getting much closer... Shucks.
Thanks for the tips. I will store them away, in case I need this info with the M8. The 5D will leave my hands in the next month or so.
Edited by carstenw on Dec 21, 2006 at 01:26 AM GMT (Reason: brain fade)
carstenw wrote:
why then does Canon only support Adobe RGB in the camera, and not ProPhoto RGB? That sucks.
The menu choice in the 5D is only relevant to JPEG out put and has zero bearing on raw. An 8-bit jpeg in ProPhoto RGB would have severe banding issues since the ProP space is so huge
As an aside, selecting ARGB in the 5D menu does have bearing on the in-camera histogram though, and thus makes a smart choice for getting the best exposure from the 5D
Now that I think about it, the worst shot I had (the red tulip shot), I developed with Lightroom.
Yes, totally predictable result -- but you'd duplicate if you converted out to ARGB from say ACR. C1 is probably smart enough to use RC with BPC as it's default conversion engine so it would likely produce a better ARGB result.
And I hear you on the Lightroom workflow, but... It is still a beta product and does not have full features yet? HOPEFULLY it will have some added -- make that real -- color management options in the final release. It would be pretty short-sighted for Adobe not to really...
I wonder about both Raw Developer and Light Zone (?) in the long run. C1 has been the same forever and I don't expect it to suddenly sprout workflow. Too bad, since I am Danish, and this would boost my nationalism
Alex wrote:
Ok, since this thread is already wild enough I will offer my speculations why Leica might have chosen a thin IR filter on the sensor. This is a little techical, and if it is not appropriate let me know, and I will delete it.
Because the hot mirror IR filters are no good for acute angles they had to use an absorption IR filters. The ammount of the absorption is determined by the distance travelled by the light ( Beer-Lambert law http://en.wikipedia.org/wiki/Beer-Lambert_law ). When the lite strikes the sensor at the normal angle the distance travelled by light in the filter is exacly the thikness of the filter. For acute angles the effective distance and the filter effect are bigger. For the 15mm lens the effective thikness of the filter in the corner is about 50% larger. It means that if the filter stops say 95% of infrared light in the center it will stop 99% in the corner. I beleive that for the wide angle lenses the magenta thing should go away in the corners. Does anybode care to verify this?
The catch with the absorption filters is that they cut into the red chanel. The green line below is the transmission function of a 2mm thick 489 IR filter. Because this effect depends on the effective thikness it becomes worse in the corners. I think that the cyan cast for noncoded wide angle lenses is exactly this - absorbed red chanel. And the correct it if the they know the focal length. And if they made the filter thick enough for IR the loss of the red chanel would have been urecoverable.
Alex thanks for this, its a more complete explanation for why the sensor filter on the M8 is thin. I'm still puzzled though because couldn't the colour shift in the corners be corrected in firmware with any lens as long as you could input its characteristics. To me this sounds like a better bet than dealing with the IR issue, is it?
Carsten - Have you tried the latest version of DPP? The software should have shipped with your 5D and you can get the update from the Canon website. As far as colours, detail and noise goes I rate it above C1 and ACR. It falls down in having little or no highlight recovery and not being as elegant in it's user interface as some.
I have tried DPP perhaps 10 months ago, but... not being a pro, I don't need to subject myself to software like DPP, C1, and so on
At the moment my philosophy is that the software with the best workflow will grow the quality needed, but the software which currently has the best quality will rest on its laurels and die.
Lightroom gives nice quality *most* of the time. I am still undecided between it and Aperture, although I currently prefer Lighroom. By the time Lightroom ships, however, Aperture may have passed it. Apple seems keen to give competition, and I do prefer Apple apps to Adobe apps, in the long run. "Adobe Photoshop Lightroom" was too much for me, I confess.
Wild-cards are Raw Developer and Light Zone. If either ends up having great workflow and database features, I would easily use them instead, but they are not there yet.
For now I will use Lightroom and when I have trouble, C1LE.
DPP produces visually pleasing images than just calibrated ACR. As a matter of fact, the files from the calibrated ACR look rather dull although there are surprisingly fine mid-tones if you watch closely.
By adding contrast or choosing mid-contrast curve, the image becomes familiar usual images with good punch at the expense of losing the mid-tone details. That led me to believe that DPP adds some artificial MSG condiment to make the images taste better.
One needs to careful with ProPhoto RGB. As the Gamut is so huge, the steps between the tones sometimes cannot be distinguished. Using 16bit is necessary.
Device profiles can "stick out" of the aRGB some times on certain saturated colors. ( kind of like stick under the tent ). It is a bad color characteristic. M8's gamut is even extreme ( according to the Lumino Landscape report ). You have to consume your data range to cover the Apexes. But you are also wasting the data range for non-spike regions. You need nice smooth convex shape
Jack Flesher:
However, if you feel you must continue to use Lightroom/ARGB
From a post by Martin Evening in Lightroom support forum:
Lightroom has an RGB space for image processing calculations that uses the ProPhoto RGB coordinates and whitepoint, but has a linear, 1.0 gamma rather than 1.8. If you are importing a file into Lightroom it does not permanently convert anything to this Lightroom RGB space, so if you import a non-raw RGB file, the calculations are done in Lightroom RGB and exported as Adobe RGB. And yes, it is safe to do so. Although there might be some concern if the original RGB file was in ProPhoto RGB and you end up with an Adobe RGB export. If you select an RGB non-raw original and choose Photo > Edit in Photoshop using Edit Original or Edit a Copy, the file will open in Photoshop in the original RGB space....Show more →
Harvey: Primarily she is saying is that LR is not going to destroy the image by permanently converting it with its current ARGB tag as it simply applys a recalculation (mapping) into ARGB, not a conversion into it. (Mapping is a non-destructive process, while conversion is destructive.)
That it uses the Photo RGB co-ordinates and white point means that the space they are working in behind the scenes is essentially ProPhoto. ProP is sufficiently large so as not destroy an image captured in any other space since ProP is already larger than any current digital imaging device can capture.
However, her comment, "...there might be some concern if the original RGB file was in ProPhoto RGB and you end up with an Adobe RGB export..." is very telling! What she probably should have said to be more precise is, "there might be some concern if the original RGB file was in A SPACE LARGER THAN Adobe RGB and you export it from LightRoom"! This is in fact precisely why Carsten is having the issues with his 5D reds after using LR to convert...
That makes sense if one starts with a larger space and winds up with a smaller one.
I have had no problems with exporting in ProPhoto space from LR to PS.
After trying CS3 beta applying some camera calibrations in ACR 4, the 5D red issues have been minimal. ACR is set to use 16 bit ProPhoto conversion when sending to PS.
The calibration I am using is modified from Thomas Fors Script results.
This is OT from M8, maybe a separate thread for this general topic as it applies to optimizing for specific camera models would be good.
Harvey Moore wrote:
I have had no problems with exporting in ProPhoto space from LR to PS.
And you won't if your original capture device gamut falls within the Adobe RGB space OR the image you capture has no colors that fall outside Adobe RGB...
As this relates to the 5D, you will only have issues in the high yellows (not very commonly occuring in nature) and very saturated reds -- reds like those that we see in bright red Danish tulips backlit by the sun
ACR4 won't give you trouble in ProPhoto because it is fully color aware where LightRoom isn't (yet).
