If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Display Modes |
#11
|
|||
|
|||
On Mon, 28 Feb 2005 07:29:11 -0800, paul wrote:
Yes I just typed "dcraw_c.bat" and it complained that BATCHCONVERT is not recognized as an internal or external command, operable program or batch file. goto was unexpected at this time. Hmmm, BATCHCONVERT is a label inside the cdraw_c.bat. It works perfectly in Win98se, no need to type anything, I can start the conversion from Irfanview or fPicasa2 or even from Windows Explorer. I'm talking about where it comes out looking almost black then you have to apply adjustments to recover. It needs to be 16 bit format. It's an unusual technique for creating B&W images with better dynamic range with high contrast images. Linear image data does NOT come out as "almost black", it just that you are not viewing them correctly, please see more at http://www.aim-dtp.net/aim/technique..._raw/index.htm Yes, correct (as long as the strings.com is version 2.5) Ah ha, I has version 1. The link on your page was not responding at the time. Sure needs the strings v2.5, I tried the d/l link just now and it is OK. I will look at the license of the freeware strings utility, it could be possible for me to put it available from my site. Timo Autiokari |
#12
|
|||
|
|||
On Mon, 28 Feb 2005 07:29:11 -0800, paul wrote:
Yes I just typed "dcraw_c.bat" and it complained that BATCHCONVERT is not recognized as an internal or external command, operable program or batch file. goto was unexpected at this time. Hmmm, BATCHCONVERT is a label inside the cdraw_c.bat. It works perfectly in Win98se, no need to type anything, I can start the conversion from Irfanview or fPicasa2 or even from Windows Explorer. I'm talking about where it comes out looking almost black then you have to apply adjustments to recover. It needs to be 16 bit format. It's an unusual technique for creating B&W images with better dynamic range with high contrast images. Linear image data does NOT come out as "almost black", it just that you are not viewing them correctly, please see more at http://www.aim-dtp.net/aim/technique..._raw/index.htm Yes, correct (as long as the strings.com is version 2.5) Ah ha, I has version 1. The link on your page was not responding at the time. Sure needs the strings v2.5, I tried the d/l link just now and it is OK. I will look at the license of the freeware strings utility, it could be possible for me to put it available from my site. Timo Autiokari |
#13
|
|||
|
|||
Hi Bart,
I've uploaded a crw for the comparison: http://www.aim-dtp.net/aim/digicam/dcraw/index.htm Timo Autiokari |
#14
|
|||
|
|||
wrote in message oups.com... SNIP I use dcraw myself and have noticed the aliasing artifacts you mention. Do have pointers to postprocessing algorithms that might help with the problem? It is best handled at conversion time itself, so it should be part of the converter software. Postprocessing the converted (and perhaps even Gamma adjusted) result will only lose information. I don't have specific pointers, although http://homepages.inf.ed.ac.uk/rbf/HIPR2/crimmins.htm#1 may provide a sense of direction. I'd personally prefer some aproach that is based on Standard Deviation rather than absolute difference. Standard deviation would come closer in appearance to Photon shot noise, which is a driving force (and physical limitation) in high signal to noise digital imaging. Bart |
#15
|
|||
|
|||
wrote in message oups.com... SNIP I use dcraw myself and have noticed the aliasing artifacts you mention. Do have pointers to postprocessing algorithms that might help with the problem? It is best handled at conversion time itself, so it should be part of the converter software. Postprocessing the converted (and perhaps even Gamma adjusted) result will only lose information. I don't have specific pointers, although http://homepages.inf.ed.ac.uk/rbf/HIPR2/crimmins.htm#1 may provide a sense of direction. I'd personally prefer some aproach that is based on Standard Deviation rather than absolute difference. Standard deviation would come closer in appearance to Photon shot noise, which is a driving force (and physical limitation) in high signal to noise digital imaging. Bart |
#16
|
|||
|
|||
Timo Autiokari wrote:
paul wrote: I just typed "dcraw_c.bat" and it complained that BATCHCONVERT is not recognized as an internal or external command, operable program or batch file. goto was unexpected at this time. Hmmm, BATCHCONVERT is a label inside the cdraw_c.bat. It works perfectly in Win98se, I did try setting the compatibility mode from winxp to win98 without any change. no need to type anything, I can start the conversion from Irfanview or fPicasa2 or even from Windows Explorer. I tried again at a DOS command line in winxp with: "dcraw_c.bat mypicture.nef" and got only the last part of the error: "goto was unexpected at this time" I don't know if it works that way, previously I had simply called the ..bat file without specifying a raw file. I also tried again by right-clicking a raw file & selecting Open-With browsing to dcraw_c.bat & grabbed a screen shot before it blinked away. There are no error messages, it simply shows the credits & closes. I've had trouble with other bat files blinking away like that before. I don't know much about DOS batch. I'm talking about where it comes out looking almost black then you have to apply adjustments to recover. It needs to be 16 bit format. It's an unusual technique for creating B&W images with better dynamic range with high contrast images. Linear image data does NOT come out as "almost black", it just that you are not viewing them correctly, please see more at http://www.aim-dtp.net/aim/technique..._raw/index.htm Here's an example of what I'm talking about: http://www.luminous-landscape.com/tu...nversion.shtml I couldn't find any documentation for dcraw command options other than a few FAQ's. I shoot in Adobe RGB but maybe that isn't being read and isn't happening, you say it should be Linear & it looks like I need a profile for photoshop to read that. I use a Nikon D70, not the D60 that you show. Still I think we are talking about different types of "linear conversion". (as long as the strings.com is version 2.5) Your link worked just now & I ran the above tests with the new version. Overseas sites sometimes don't connect from california I've noticed it's not unusual, nobody's fault. I appreciate your efforts. |
#17
|
|||
|
|||
paul wrote:
Timo Autiokari wrote: I'm talking about where it comes out looking almost black then you have to apply adjustments to recover. It needs to be 16 bit format. It's an unusual technique for creating B&W images with better dynamic range with high contrast images. Linear image data does NOT come out as "almost black", it just that you are not viewing them correctly, please see more at http://www.aim-dtp.net/aim/technique..._raw/index.htm Here's an example of what I'm talking about: http://www.luminous-landscape.com/tu...nversion.shtml I couldn't find any documentation for dcraw command options other than a few FAQ's. I shoot in Adobe RGB but maybe that isn't being read and isn't happening, you say it should be Linear & it looks like I need a profile for photoshop to read that. I use a Nikon D70, not the D60 that you show. Still I think we are talking about different types of "linear conversion". I played some more at the command prompt with plain dcraw. Simply type dcraw to get a list of the options Doh! Your settings: set DCRAWcmd=-v -b 4 -3 -b 4 increases brightness to counteract the dark look of a linear conversion that I mentioned. If I don't use this option, I can adjust the levels in photoshop and get a more robust looking, less clipped histogram than the Adobe RAW plugin. Also the less agressive antialiasing filter in dcraw gives a sharper but noisier image which has advantages. I did play with the hybrid approach as well since I have no profile for the linear conversion the dcraw colors came out odd as well as the color bayer artifacts mentioned above. One strange thing is the dcraw conversion frames the image up about 20 pixels (portrait shot) so it doesn't match the adobe version. Very very odd! |
#18
|
|||
|
|||
Hi Paul,
On Tue, 01 Mar 2005 09:18:31 -0800, paul wrote: Your settings: set DCRAWcmd=-v -b 4 -3 -b 4 increases brightness to counteract the dark look of a linear conversion that I mentioned. Aha, I did not understand that it was _this_ issue. I scale the 12-bit output of my D60 up using DCRAW because there will be less quantization (or there are more different RGB values) than how Photoshop Levels does it. since I have no profile for the linear conversion the dcraw colors came out odd as well as the color bayer artifacts mentioned above. Yes, it is best to have profile(s). My understanding is that the default color space that DCRAW does the linear conversion to is Widegamut, D65 gamma 1.0 profile. This is the same as my AIMRGBpro profile, available at: http://www.aim-dtp.net/aim/download/aim_profiles.zip So daylight/sunlight shots should appear rather nicely if you just Assign the AIMRGBpro profile. One strange thing is the dcraw conversion frames the image up about 20 pixels (portrait shot) so it doesn't match the adobe version. Very very odd! It is so with D60 conversion also, but just a few pixels, larger in both x and y dimensions. I believe that what DCRAW outputs is the real sensor size. I'm sorry I am not able to help you with the batch at this time, as soon as I have the possibility to test it on an XP machine I will take a good look into this. But in addition of being logged as an administrator I believe you still need to give the proper rights for the dcraw_c. Timo Autiokari |
#19
|
|||
|
|||
"Timo Autiokari" wrote in message ... Hi Bart, I've uploaded a crw for the comparison: http://www.aim-dtp.net/aim/digicam/dcraw/index.htm Okay. Before I give you the links to my conversion results, allow me to make a introductory comment about the file. This file should be almost perfect for DCRAW, because there is almost nothing in the file that shows focused straight edges. The multicolored high contrast edges are the Achilles heel of DCRAW. However, even in this file the rainbow like colors are present in the ring edges of the bridle. The good news is that the colored artifacts don't have to show as prominently if they were worsened by sharpening. Just apply the sharpening on a duplicate layer and use a Luminosity blending mode. Now for the results, I've prepared the following: 1. From Rawshooter essentials, http://www.xs4all.nl/~bvdwolf/temp/Canon-RSE.jpg . The left window is unsharpened from RSE-2005, and the right window is the same file but with USM 170 / 0.3 / 0 . I tried to achieve a similar amount of sharpening as with your DCRAW examples. 2. From Adobe Camera Raw, http://www.xs4all.nl/~bvdwolf/temp/Canon-ACR.jpg . The left window is unsharpened from Adobe Camera Raw, and the right window is the same file but with USM 500 / 0.3 / 0 . Again, I tried to achieve a similar amount of sharpening as with your DCRAW examples. Different sharpening methods will also produce different looking results. To me, the ACR example shows a very slight advantage to both others, due to less posterized shadows than DCRAW (but that posterization has probably more to do with the profile you used), and the slightly better shadow definition than RSE. Do note that in trying to match the output in color and tonality of the DCRAW file, things may have been lost. Straight conversion with these converters produces slightly different output, at which point it becomes a matter of taste and it makes a direct comparison more difficult. Different files/subjects will also produce different results. Nevertheless, DCRAW is a fine converter if it weren't for the multicolored artifacts on high contrast edges, and spotty thin lines. It would seem useful to search for a solution to solve that. A simple method for profiling/color matching would be welcome as well. The free "Rawshooter essentials 2005" is a very good tool, if the current issues with other platforms than P4/Win XP are solved. The workflow is well thought through, and the stand alone application allows good sharp output, with very little color fringing (it seems to suppress chromatic aberration to a certain extent). A little more control in a payware version would be appreciated. It is also possible to produce aliasing artifacts with extremely fine detail, it is usually low in color moiré but visible as luminance moiré. Adobe Camera Raw is an all purpose converter and offers lots of control in a color-managed environment. It produces quality output, but it requires Photoshop CS for a full capability version of the Raw converter. Bart |
#20
|
|||
|
|||
Hi Bart,
Thank you! Is it OK to put all the results side by side on the dcraw page at my site? This file should be almost perfect for DCRAW, because there is almost nothing in the file hat shows focused straight edges. The multicolored high contrast edges are the Achilles heel of DCRAW. I agree that the edges in DCRAW conversion are not as soft as they are from what ever other converter but really these artifacts that you refer to do not show on the CRT at 100% zoom (at actual pixels size). A CRT has about 100ppi so this means a printed image e.g. from my D60 (6MP) at the size of 50cm by 75cm (20 inch by 30 inch). Sure, if I need to resample up from that then there will be a need to address to this issue but there are several things that can be done. Do you truly see rainbow like colors at the edges of those rings? At what magnification? Also, where exactly do you see posterized shadows in my DCRAW version??? I can not see any. From these the DCRAW conversion is clearly the best (most sharp), ACR is the second best but definitely more soft. The RSE conversion seems to have also other problems in addition to the softness (e.g. blotchy skin on the neck). The Canon software clearly is the worst (extremely soft), one has to remember that it does not apply any sharpening in linear the conversion mode so for a fair comparison one actually should sharpen them much more ...but that does not help at all, just gives more sharpening artefacts. The color-managment is another kind of issue, I prefer DCRAW because with it I have full control over that. I have currently 12 profiles for my D60 (for different illumination situations) but the common 4 or 5 that there are in commercial/bundled converters is quite enough normally. It is not a very big task at all to create these camera profiles but the commercial products sure are easier to take into use in regars to color-management, certainly not equally accurate, but more easy. And of course the Windows GUI is an issue but the batch file now makes things very very easy. On the other hand DCRAW workflow (once the initial setup is ready) is very simple, straightforward and clear / easily understandable, you get the raw image data into Photoshop, engage the color-management and then finalize the image fully there using those familar editing operations ...and there are quite many operations to choose from in Photoshop. For me these "full-featured" converters are like blackboxes ...and Photoshop acts as just as a mere "SaveAs engine" or a JPEG converter behind them. One thing btw should be very clear now: The conversion algorithm have a paramount effect on image quality. E.g. one can find many Web pages that focus on the lens quality, inspecting differences between the lenses that are just a tiny fraction of that softness that a poor conversion algorithm produce. Really there is no benefit in buying an extremely expensive "sharp" lens in case one is using a "soft" converison tool. Timo Autiokari |
Thread Tools | |
Display Modes | |
|
|