 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
| PhotoKB HomeDiscussion Groups Digital Photography Digital PhotoDSLR CamerasZLR CamerasPoint & Shoot CamerasFilm Photography 35 mmLarge FormatMedium formatDarkroomFilm and LabsOther EquipmentPhoto Technique Nature PhotographyPeople PhotographyTechnique GeneralGeneral Photo Topics General TopicsAustralian PhotographyUK PhotographyDirectoryPhoto Clubs |
Photo Forum / Digital Photography / DSLR Cameras / July 2006Save over-exposed photo
I took some photos a couple weeks ago at my cousin's wedding. There is one shot I would really like to save. I was setting up & taking some test shots while everyone assembled for photos on a little bridge over a small pond. This photo was one of those test shots, where I was trying to get my exposure correct but hadn't quite got things right yet. I really like the (completely accidental) composition. I'd be very thankful for any help or pointers trying to save the photo. My Nikon raw file is here: <http://thepierianspring.org/wedding.nef> (6 MB) I bought a copy of Photoshop CS2 early this week and have made a few attempts, but I'm still trying to learn it. Here is my best attempt so far which tries to save it by stylizing it a bit (added "density" with the duplicate layer and multiply method; applied crosshatch filter; created soft focus effect; and added frame): <http://thepierianspring.org/wedding.png> (8 MB) <http://thepierianspring.org/wedding.jpg> (1 MB) Thanks for any suggestions, Randy. From a distance the photo may have some redeeming features but I think if you look closely some of the faces are screwed up and things like that so in my view wouuld not be worth the effort. The highlights are badly blown and have lost all detail, possibly the best idea might be to convert to B&W and then hand colour. >I took some photos a couple weeks ago at my cousin's wedding. There is one >shot I would really like to save. I was setting up & taking some test shots [quoted text clipped - 19 lines] > Thanks for any suggestions, > Randy. > From a distance the photo may have some redeeming features but I think if > you look closely some of the faces are screwed up and things like that so in > my view wouuld not be worth the effort. The highlights are badly blown and > have lost all detail, possibly the best idea might be to convert to B&W and > then hand colour. I'm not so sure about this. There was someone in this forum who extracted so much detail out of an apparently blown image of a chinese calligraphist (I think it was) at work, it was simply magical. I seem to remember it involved channel separation of some kind, but never did get to hear the details of how he accomplished it. Perhaps he would kindly repeat the party trick again for us with this one? >> From a distance the photo may have some redeeming features but I think if >> you look closely some of the faces are screwed up and things like that so [quoted text clipped - 8 lines] > details of how he accomplished it. Perhaps he would kindly repeat the > party trick again for us with this one? Let the challenge begin! > >> From a distance the photo may have some redeeming features but I think if > >> you look closely some of the faces are screwed up and things like that so [quoted text clipped - 10 lines] > > Let the challenge begin! You know, that's a great idea for all the PS wonks out there. Maybe the OP can set up a "Reward Fund" or Certificate or something for the person who posts the best correction/fix of the photo. But the winner (after the competion) would have to describe exactly what he/she did to fix it. That's keep a slow NG week interesting. >> From a distance the photo may have some redeeming features but I think if >> you look closely some of the faces are screwed up and things like that so in [quoted text clipped - 5 lines] >extracted so much detail out of an apparently blown image of a chinese >calligraphist (I think it was) at work, it was simply magical. That may have been me (someone else did something, too). The blue channel in incandescent light is difficult to blow out, because it is two stops weaker than the green and red. In this particular (wedding.nef) )case, the bright clothing and the woman on the left's face are all completely blown in all three channels, so there is no luminance info.  Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > That may have been me (someone else did something, too). The blue > channel in incandescent light is difficult to blow out, because it is > two stops weaker than the green and red. Ah, yes. That was it, thanks. > In this particular (wedding.nef) )case, the bright clothing and the > woman on the left's face are all completely blown in all three channels, > so there is no luminance info. OK, so I am beginning to understand this better. So long as one channel contains luminance data, then it may be possible to reconstruct an image such as this, provided the detail itself (such as the calligraphy in the other image you worked on) is monochromatic? >OK, so I am beginning to understand this better. So long as one channel >contains luminance data, then it may be possible to reconstruct an image >such as this, provided the detail itself (such as the calligraphy in the >other image you worked on) is monochromatic? The data is monochromatic because there is only one color channel. It's a result of the clipping, not a feature of the subject. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >> OK, so I am beginning to understand this better. So long as one channel >> contains luminance data, then it may be possible to reconstruct an image [quoted text clipped - 3 lines] > The data is monochromatic because there is only one color channel. It's > a result of the clipping, not a feature of the subject. Thanks John. My mind works in mysterious ways, few wonders to perform. ;-) >>> OK, so I am beginning to understand this better. So long as one channel >>> contains luminance data, then it may be possible to reconstruct an image [quoted text clipped - 5 lines] > > Thanks John. My mind works in mysterious ways, few wonders to perform. ;-) So I was right? Well in this case at least! > So I was right? Well in this case at least! You may have reached the right conclusion, but I don't see any convincing explanation as to how you achieved it? :-P I would still perhaps enjoy delving a little more into how John managed the technical feat of dealing with the calligraphist image, as I'm sure this is one which could reap rewards for many of us here coping with extreme dynamic range images. I don't see that sort of information in the usual sources, and it surprises me. Or maybe I am not as well 'connected' as I perhaps thought. ;-) >I would still perhaps enjoy delving a little more into how John managed >the technical feat of dealing with the calligraphist image, as I'm sure >this is one which could reap rewards for many of us here coping with >extreme dynamic range images. It's not much of a feat to get the least clipped RAW color channel from the RAW file. Just load it into IRIS (with the load RAW menu item), set the thresholds to 4095 and 0, then open the command window and type "split_cfa a b c d" and then alternately "load b", "load c", "load d", to see which channel is least clipped. The colors that correspond to a, b, c, and d depend on the camera. Once you have loaded the least clipped one, you can then save as a 16-bit TIFF or an 8-bit JPEG or BMP. If you're saving for photoshop in 16-bit, you need to go into the processing menu and multiply the image by 2 not to get posterized in photoshop, before saving it. The 16-bit TIFF saves literal values; the 8-bit options save it as it appears on the screen, with the threshold settings. Some cameras, like Canons, don't have black at zero in the RAW data, so you have to know what black is and subtract it from the image. Black is 128 for most Canons at most or all ISOs (it is more variable with the older cameras). Black is 0 for most Nikons. Don't forget that black varies if you multiply the image, so if you subtract after multiplication you have to multiply the number subtracted as well. Another method would be to click on the camera icon on the toolbar of IRIS, and select the camera model at the bottom of the tab, and then use the menu command "convert a CFA image", which will take the 3 color channels and interpolate them into full-MP channels. You can save this out and look at the individual color channels in photoshop. It is easier to subtract the blackpoint in IRIS before saving out than trying to do it in photoshop; photoshop does not have any kind of precision when working with the Levels tool in 16-bit mode. >I don't see that sort of information in >the usual sources, and it surprises me. Or maybe I am not as well >'connected' as I perhaps thought. ;-) Most of the few people who know much about how RAW data works are developers, so obviously, they are not going to talk much about the details, because they don't want to aid competition. We are living in the dark ages of digital photography, IMO. This is really simple stuff, once you grasp the issues of blackpoint and color balance. It's really a shame that 99.99% of discussion about DSLR imaging issues are based on JPEGs, when RAW data tells a much more direct and unadulterated story. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >>I would still perhaps enjoy delving a little more into how John managed >>the technical feat of dealing with the calligraphist image, as I'm sure [quoted text clipped - 45 lines] > direct and unadulterated story. > ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< What an interesting discussion this has turned into. R. > What an interesting discussion this has turned into. > > R. Absolutely. I remember when John was kindly doing his magic with what I refer to as the 'Calligraphist' image, and very much wanted to pursue this in further detail, such as he has very kindly just done. However, I unfortunately had other matters intervene at the time, and had to drop it. I now know about the existence of IRIS. :-) > We are living in the dark ages of digital photography, IMO. This is > really simple stuff, once you grasp the issues of blackpoint and color > balance. It's really a shame that 99.99% of discussion about DSLR > imaging issues are based on JPEGs, when RAW data tells a much more > direct and unadulterated story. Thanks very much indeed for this fascinating information, John. It is kind of you to go to such lengths. I have now obtained a copy of IRIS, and will practise on this presently together with a copy of your instructions. As to the matter of RAW itself, I too would very much like to encourage as many as possible to engage in discussing this format more. I assume it's extremely close relationship with DSLR's isn't going to tread on any toes in here? :-) > Thanks very much indeed for this fascinating information, John. It is > kind of you to go to such lengths. I have now obtained a copy of IRIS, > and will practise on this presently together with a copy of your > instructions. 8< Oh dear. We fall at the first hurdle. It won't read my cams (.dcr) RAW files. Sod it. Back to the drawing board. Let's hope the RAW Magick team come up with something special in their new RMeX software due for alpha testing within the next couple of days. >> Thanks very much indeed for this fascinating information, John. It is >> kind of you to go to such lengths. I have now obtained a copy of IRIS, [quoted text clipped - 7 lines] > come up with something special in their new RMeX software due for alpha > testing within the next couple of days. This is where it would be good to have the program read DNG files. >> Thanks very much indeed for this fascinating information, John. It is >> kind of you to go to such lengths. I have now obtained a copy of IRIS, [quoted text clipped - 7 lines] >come up with something special in their new RMeX software due for alpha >testing within the next couple of days. DCRAW might work, but I don't use it enough to be able to tell you what settings to use to get a RAW-like output. You want no white balance and no conversion to a color space. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > I took some photos a couple weeks ago at my cousin's wedding. There is > one shot I would really like to save. I was setting up & taking some [quoted text clipped - 19 lines] > Thanks for any suggestions, > Randy. Doen't look bad if you take the exposure done to -3.00, drop the contrast way down -25 - -30 and shadow to 0. I think the woman on the left is a loss, but the rest doesn't look bad at all. Don't screw arround with cross hatches etc, you also have too much sharpening (halos around heads). Don't sharpen and soft focus. If you have PSCS2 use Smart Sharpener, much better, and controllable than Unsharp Mask. Tom >>I took some photos a couple weeks ago at my cousin's wedding. There is >>one shot I would really like to save. I was setting up & taking some [quoted text clipped - 26 lines] > (halos around heads). Don't sharpen and soft focus. If you have PSCS2 > use Smart Sharpener, much better, and controllable than Unsharp Mask. In CS1 ACR I dropped every slider to the left including the saturation. The colors are probably hopeless but I got a nice B&W image. The other thing I did was fiddle the WB temperature to 4000 till I got the most detail/least posterization out of the lady on the left's face zoomed in and dropped the tint WB slider to -12. Same routine in RSE gave a smoother lower contrast image. Then I pasted the RSE version over the ACR & set the RSE opacity to 33% & it looks pretty decent although better results could be achieved by selectively masking these versions together depending on which looks best in a particular area. There is a horrible pixellated grit on the man's white jacket in the ACR version for example. I found that any sharpening at all just made a mess of things and I don't think your noise reduction approach looks realistic although I'm amazed you got the colors looking OK so that could be overlaid in color only mode or even used as is for the background but the people & faces should be more realistic. Here it is as I described above: http://www.edgehill.net/temp/wedding-web.jpg SignaturePaul Furman http://www.edgehill.net/1 Bay Natives http://www.baynatives.com >>> I took some photos a couple weeks ago at my cousin's wedding. There is >>> one shot I would really like to save. I was setting up & taking some [quoted text clipped - 45 lines] > > http://www.edgehill.net/temp/wedding-web.jpg Thanks Tom and Paul. I'll play with these some when I get home. IIRC, none of my other photos were taken from this same point of view. This shot was taken from a steep embankment far enough away that I didn't have enough zoom to get any closer shots, so my others were taken closer up and at a lower level; where the people obscure much of the grassy field in the background. What I'm wondering is if I might still be able to clone some areas from other photos to retouch some of the background. That field behind the pond is really a full field of green grass. That would help a good bit. Is it possible in Photoshop to clone from another photo? Sure it is; I'll just have to figure out how. Maybe I can use the vanishing point tool to clone parts of the bridge over the bad spots there. Not sure about the girl on the left yet... How to get rid of the blue cast in the shadows on the bride's dress and the guy beside her? Maybe I'm spending too much time on this one, but for some reason, I really like this shot. It reminds me of an older style painting or something. Thanks much for all the comments and help, Randy. >I took some photos a couple weeks ago at my cousin's wedding. There is >one shot I would really like to save. I was setting up & taking some [quoted text clipped - 5 lines] >I'd be very thankful for any help or pointers trying to save the photo. >My Nikon raw file is here: You can't save it. All three RAW channels are severely clipped. The red channel is the least clipped, and even that is solidly clipped in all of the bright clothing. Had only the red channel remain unclipped, you could have had clothing detail for white clothes from a converter capable of doing so. The faces themselves are unclipped, except for the woman on the extreme left. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > I took some photos a couple weeks ago at my cousin's wedding. There is > one shot I would really like to save. I was setting up & taking some [quoted text clipped - 19 lines] > Thanks for any suggestions, > Randy. Here is my attempt. As others said, all channels are clipped. I found the blue channel to be the best. The method I used: ImagesPlus 2.75: 1) convert from raw to linear image without bayer interpolation. 2) extract each channel and write as a 16-bit tif. 3) in photoshop use curves and shadow/highlight to improve the image as much as possible. 16-bit really was not needed. http://www.clarkvision.com/tmp/wedding_cfa_B1b.jpg It is black and white only, and full resolution (even though 1520x 1007 pixels--that is how many blue pixels there are). Roger Roger, you definately are THE MAN! >> I took some photos a couple weeks ago at my cousin's wedding. There is >> one shot I would really like to save. I was setting up & taking some test [quoted text clipped - 34 lines] > > Roger In message <44A89FF8.3080603@qwest.net>, "Roger N. Clark (change username to rnclark)" <username@qwest.net> wrote: >Here is my attempt. As others said, all channels are clipped. >I found the blue channel to be the best. The method I used: [quoted text clipped - 3 lines] >to improve the image as much as possible. 16-bit really was >not needed. >http://www.clarkvision.com/tmp/wedding_cfa_B1b.jpg >It is black and white only, and full resolution (even though 1520x >1007 pixels--that is how many blue pixels there are). Something is not right with your workflow; your image clips about 1.5 stops below where the RAW blue channel clips: http://www.pbase.com/image/62900177/original I'm getting 255 in your image where the RAW is only about 1300 out of 4095. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > In message <44A89FF8.3080603@qwest.net>, > "Roger N. Clark (change username to rnclark)" <username@qwest.net> [quoted text clipped - 20 lines] > I'm getting 255 in your image where the RAW is only about 1300 out of > 4095. I'm very glad you said that John. I was having trouble believing my own eyes, especially as Roger is somewhat of a dab hand in these matters by all accounts. Even I could get your result using IRIS!! lol >In message <44A89FF8.3080603@qwest.net>, >"Roger N. Clark (change username to rnclark)" <username@qwest.net> >wrote: >>Here is my attempt. >Something is not right with your workflow; your image clips about 1.5 >stops below where the RAW blue channel clips: >http://www.pbase.com/image/62900177/original >I'm getting 255 in your image where the RAW is only about 1300 out of >4095. I just looked a little closer; the channel I posted seems to be red; that is the least clipped channel. Yours does seem to be blue, which has more clipped areas, but your workflow still clips it at about 3000, which generally becomes about 4095 in the blue channel with daylight WB, so maybe you have some unintended WB in your image. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >>In message <44A89FF8.3080603@qwest.net>, >>"Roger N. Clark (change username to rnclark)" <username@qwest.net> [quoted text clipped - 15 lines] > which generally becomes about 4095 in the blue channel with daylight WB, > so maybe you have some unintended WB in your image. John, That may be. I used ImagesPlus 2.75 and while the converter for Canon cameras has a lot of options, the Nikon converter does not, only standard, and no CFA. So this difference seems to be an ImagsPlus limitation. I'll post to the IP group and see what they say. Roger |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.