Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
Rob wrote: > Dave wrote: > >It seems to me from eyeball guessing that my LS-30 is resolving grain > >in 100 ISO films at roughly 40-80% distortion, which looks pretty bad > >on the monitor at 100% view. 800 speed color neg film does much > >better at what I would guess to be roughly 25% distortion. > > I presume you're comparing 100ASA print film with 800ASA print film? > (as opposed to 100ASA slide film) > Out of interest, exactly which brand/types of film are you using? > How did you judge the distortion? Compared to what? I don't see significant differences in grain at the print level between 100 speed negs and chromes, and print level is all I really care about. I'm scanning a variety of films from my files etc, but these days I tend to shoot mostly Fuji Provia 100, Astia 100, 64T, NPS 160, and NHG 800. (I guess I like Fuji :) My judgement is completely subjective and therefore probably not worth too much. (Take with a large "grain" of halide :) I judged the distortion by comparing the grain on monitor at 100% to how I think it would look with no distortion, and to tonal areas in the same scan with less aliasing distortion. My ideas about how grain looks are formed by seeing grain magnified in various ways over the years. Generally, the grain in least aliased areas of 800 speed neg film looks pretty close to no distortion with LS-30 scans, to my eye. > >The silver lining to this cloud is aliasing distortion (with the > >LS-30) looks worse on screen than print, IMO. > > I'd agree with this, although in some cases the amount of distortion > (aliasing, grain, whatever you want to call it) is ugly at much lower > print sizes. I have a panoramic print on my wall at work on Epson > Panoramic Photo Paper - the ocean and beach looks fine, but the sky > has an ugly brown discolouration caused by grain aliasing. The film > was either Fuji Superia 100 or Reala. I'm not sure how aliasing distortion could cause color shift, but since aliasing becomes greater at certain tonal transitions you may be seeing the additive effect of two problems overlapping. Does your printer profile posterize blues? > Generally however, the printer does tend to be more fogiving than the > monitor - the "grain" usually ends up less intense in a print. Jon Cone suggests to avoid overemphasing grain when sharpening scans (I suppose equally true for aliasing distortion), never use a radius setting higher than 0.8. That's the way I tend to work editing Agfa T-2500 scans, but I usually turn on ICE and sharpening together in NikonScan 2.5, and it seems about right to me. Dave
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
Dave wrote: >I don't see significant differences in grain at the print level >between 100 speed negs and chromes, and print level is all I really >care about. Really??! In the scans I see a huge difference between say Superia 100 and Sensia II 100. There's a *much* bigger difference when you go to Provia 100F. There is no apparent grain with Provia 100F. Sensia II shows little grain. Superia 100 or Reala shows a lot of ugly grain in highlights - particularly skies. I haven't done enough large prints to know for sure, but from the evidence I have there's a big difference between the quality of prints from slides compared to prints from negs. *However* one thing I don't recall you mentioning was what program you're using to scan with. Are you using Nikonscan? If so, I'm not suprised you have less problems with grain in the sky. Vuescan seems to increase grain in highlight areas of negs, and earlier versions put a distinct orange/brown tint in the image compared to Nikonscan. The detail in the skies tend to "blow out" in Nikonscan with the LS30 since it only works with 8 bit data - this has the side effect of reducing apparent grain in the sky. Unfortunately Nikonscan is useless for me since I get jaggies with it, so I have to use Vuescan. I may be able to "improve" things a little by deliberately adjusting the white point, but I don't want to lose too much sky detail. > I'm scanning a variety of films from my files etc, but > these days I tend to shoot mostly Fuji Provia 100, > Astia 100, 64T, NPS 160, and NHG 800. (I guess I like Fuji :) I mostly use Fuji as well. The only non-fuji film I've used much in the last ten years is Kodak T400CN. At the moment I generally use Superia 100 or Provia 100F. >My judgement is completely subjective and therefore probably not worth >too much. (Take with a large "grain" of halide :) I judged the >distortion by comparing the grain on monitor at 100% to how I think it >would look with no distortion, and to tonal areas in the same scan >with less aliasing distortion. Ah. I was wondering whether you were comparing it to grain in a photographic 10x8 print or something. > My ideas about how grain looks are formed by seeing grain > magnified in various ways over the years. The joys of experience. :) I'm not so fortunate. >Generally, the grain in least aliased areas of 800 speed neg film >looks pretty close to no distortion with LS-30 scans, to my eye. Meaning the grain is real? To me, Fuji 800 looks very grainy when scanned. >I'm not sure how aliasing distortion could cause color shift, but >since aliasing becomes greater at certain tonal transitions you may be >seeing the additive effect of two problems overlapping. Does your >printer profile posterize blues? I'm not using the same printer anymore, but regardless - the colour shift was in the scanned image. See my comments above about Vuescan and negs. Vuescan does seem to highlight odd colours in blue skies for some reason when scanning negs. This may be an old problem though - I'd have to try scanning the same panoramic frame again with the current incarnation of Vuescan. Ed has made a LOT of changes to the colour transformations since I made the original scan. Sky grain is still an issue though. > Jon Cone suggests to avoid overemphasing grain when sharpening scans > (I suppose equally true for aliasing distortion), never use a radius > setting higher than 0.8. This is interesting. I'll have to check my unsharp mask settings. Having said that - I don't always sharpen scans prior to printing. I imagine sharpening would be more important when printing to larger sizes as you are. The largest I can print at the moment is A3+, and I have yet to try it. I've only printed to something a little smaller than A3. >NikonScan 2.5, and it seems about right to me. I guess this answers my question about what program you use. :) Rob Rob Geraghty [EMAIL PROTECTED] http://wordweb.com
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
> Dave wrote: > >I don't see significant differences in grain at the print level > >between 100 speed negs and chromes, and print level is all I really > >care about. > > Really??! In the scans I see a huge difference between say Superia 100 > and Sensia II 100. There's a *much* bigger difference when you go to Provia > 100F. There is no apparent grain with Provia 100F. Sensia II shows little > grain. Superia 100 or Reala shows a lot of ugly grain in highlights - particularly > skies. I haven't done enough large prints to know for sure, but from the > evidence I have there's a big difference between the quality of prints from > slides compared to prints from negs. In general grain doesn't bother me, it's an artifact of silver halide technology I've grown to accept. I *do* want to see it in the print, and I want it to look very similar to how it would in a very good equal size analog print. How it looks on the monitor doesn't concern me really. When magnified you'll see significant differences in grain size and structure between various emulsions, yes. As a general rule, it's pretty easy to inadvertently overemphasize grain in digital, so I try to avoid that. > *However* one thing I don't recall you mentioning was what program you're > using to scan with. Are you using Nikonscan? If so, I'm not suprised you > have less problems with grain in the sky. Vuescan seems to increase grain > in highlight areas of negs, and earlier versions put a distinct orange/brown > tint in the image compared to Nikonscan. I haven't used Vuescan since figuring out how to get what I want out of NikonScan. > The detail in the skies tend to "blow out" in Nikonscan with the LS30 since > it only works with 8 bit data - this has the side effect of reducing apparent > grain in the sky. Unfortunately Nikonscan is useless for me since I get > jaggies with it, so I have to use Vuescan. I may be able to "improve" things > a little by deliberately adjusting the white point, but I don't want to > lose too much sky detail. I see some artifacts from ICE on edges of tonal contrast areas on the monitor, but again, in print it's a non-issue, so I've been able to ignore it. > > I'm scanning a variety of films from my files etc, but > > these days I tend to shoot mostly Fuji Provia 100, > > Astia 100, 64T, NPS 160, and NHG 800. (I guess I like Fuji :) > > I mostly use Fuji as well. The only non-fuji film I've used much in the > last ten years is Kodak T400CN. At the moment I generally use Superia 100 > or Provia 100F. > > >My judgement is completely subjective and therefore probably not worth > >too much. (Take with a large "grain" of halide :) I judged the > >distortion by comparing the grain on monitor at 100% to how I think it > >would look with no distortion, and to tonal areas in the same scan > >with less aliasing distortion. > > Ah. I was wondering whether you were comparing it to grain in a photographic > 10x8 print or something. > > > My ideas about how grain looks are formed by seeing grain > > magnified in various ways over the years. > > The joys of experience. :) I'm not so fortunate. Or perhaps you are . > >Generally, the grain in least aliased areas of 800 speed neg film > >looks pretty close to no distortion with LS-30 scans, to my eye. > > Meaning the grain is real? To me, Fuji 800 looks very grainy when scanned. "Real" is a relative term. Aliasing is a "reaction formation" to real grain, to varying degrees of visual distortion. If the distortion artifacts don't obliterate the grain structure completely (and to my way of looking at it they don't), then at *lower magnifications* (prints) the essential qualities of the original are more or less intact. Fujicolor 800 looks relatively grainy because it is. But it has a smooth structure, and looks great in large prints. Wouldn't shoot architecture on 35mm with it :), but for personal work I love it. Sometimes it "packs up" on smaller prints (don't know better way to describe it), to me it looks better at bigger print sizes oddly enough, but I don't mind it small either, it does what it does. So much improvement in high speed neg films in the past couple of years. I wouldn't have considered an 800 speed neg film years ago except in "emergencies". > >I'm not sure how aliasing distortion could cause color shift, but > >since aliasing becomes greater at certain tonal transitions you may be > >seeing the additive effect of two problems overlapping. Does your > >printer profile posterize blues? > > I'm not using the same printer anymore, but regardless - the colour shift > was in the scanned image. See my comments above about Vuescan and negs. > Vuescan does seem to highlight odd colours in blue skies for some reason > when scanning negs. This may be an old problem though - I'd have to try > scanning the same panoramic frame again with the current incarnation of > Vuescan. > Ed has made a LOT of changes to the colour transformations si
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
Rob wrote: > The detail in the skies tend to "blow out" in Nikonscan with the LS30 since > it only works with 8 bit data - this has the side effect of reducing apparent > grain in the sky. Unfortunately Nikonscan is useless for me since I get > jaggies with it, so I have to use Vuescan. I may be able to "improve" things > a little by deliberately adjusting the white point, but I don't want to > lose too much sky detail. The trick with the LS-30 is to hardware calibrate your monitor (PhotoCal/Monitor Spyder is great, and not too expensive), set up color management in NikonScan using the supplied scanner profile, and use the excellent NikonScan curves dialogue to tone/color correct before scanning. You'll be outputting 24 bit files to PS, but the corrections are applied in hi bit space. Then, if need be, apply tweaks with an adjustment layer before printing. Nikonscan's CM works as well as possible, with a near perfect match to the result in Photoshop. Also Nikonscan does the best color corrections out of the box of anything I've seen, on chromes and negs. And, as I noted previously, the sharpening algorithm it uses is very good. Dave
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
Dave wrote: >Nikonscan's CM works as well as possible, with a near perfect match to >the result in Photoshop. Also Nikonscan does the best color >corrections out of the box of anything I've seen, on chromes and negs. >And, as I noted previously, the sharpening algorithm it uses is very >good. Er, but as I noted previously, Nikonscan is *useless* for me as it produces jaggies. Vuescan is my only option. Yes, I really like the colour adjustment facilities in Nikonscan, but they're no good to me if the output is garbage due to a bug. Rob Rob Geraghty [EMAIL PROTECTED] http://wordweb.com
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
> Dave wrote: > >Nikonscan's CM works as well as possible, with a near perfect match to > >the result in Photoshop. Also Nikonscan does the best color > >corrections out of the box of anything I've seen, on chromes and negs. > >And, as I noted previously, the sharpening algorithm it uses is very > >good. > > Er, but as I noted previously, Nikonscan is *useless* for me as it produces > jaggies. Vuescan is my only option. Yes, I really like the colour adjustment > facilities in Nikonscan, but they're no good to me if the output is garbage > due to a bug. > > Rob Do you mean jaggies are all through the image, or along the edges? Dave
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
I also would like to put a word of support for Nikonscan here. I use LS2000 and Nikonscan 2.5.1. I have tried Vuescan but just can't get it to do anything better than Nikonscan (EXCEPT reduce jaggies) so I continue to use Nikonscan. There has been a lot of negative discussion about Nikonscan - I really cannot see why people bag it so much. I get predictable output and generally excellent colour 98% of the time from negs. I don't do much slides, but they were fine too. I certainly get better results colour-wise than I could ever get out of Vuescan, and VS was *much* slower. I do agree that the jaggies is a real problem, and have not been impressed by the results of my email discussion with Nikon USA about this. I am about to send mine back for "repair" re jaggies - I have little hope but will report how it is dealt with. (remember this is in Australia). I also have troubles with focus depth of field, but that is not a software problem. re the blown highlights / loss of sky detail comment, my technique to avoid this is as follows... IME NIkonscan default auto settings cut off too much at the high end (and maybe the shadows end too), so I use the option - Scanner extras / prescan mode / low contrast neutral. This means I get the whole range of a neg into the histogram. At 16-bit there is no problem re-expanding it to get the cut off at the actual tips of the white and black points, then I can do whatever contrast enhancement etc I need in PS. Not sure how this translates to the LS30, but I think it is still valid. Julian At 05:01 10/04/01, you wrote: >Rob wrote: > > > The detail in the skies tend to "blow out" in Nikonscan with the >LS30 since > > it only works with 8 bit data - this has the side effect of reducing >apparent > > grain in the sky. Unfortunately Nikonscan is useless for me since I >get > > jaggies with it, so I have to use Vuescan. I may be able to >"improve" things > > a little by deliberately adjusting the white point, but I don't want >to > > lose too much sky detail. > >The trick with the LS-30 is to hardware calibrate your monitor >(PhotoCal/Monitor Spyder is great, and not too expensive), set up >color management in NikonScan using the supplied scanner profile, and >use the excellent NikonScan curves dialogue to tone/color correct >before scanning. You'll be outputting 24 bit files to PS, but the >corrections are applied in hi bit space. Then, if need be, apply >tweaks with an adjustment layer before printing. > >Nikonscan's CM works as well as possible, with a near perfect match to >the result in Photoshop. Also Nikonscan does the best color >corrections out of the box of anything I've seen, on chromes and negs. >And, as I noted previously, the sharpening algorithm it uses is very >good. > >Dave Julian Robinson in usually sunny, smog free Canberra, Australia
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
Julian wrote: > I also would like to put a word of support for Nikonscan > here. I use LS2000 and Nikonscan 2.5.1. I have tried > Vuescan but just can't get it to do anything better than > Nikonscan (EXCEPT reduce jaggies) so I continue to use > Nikonscan. There has been a lot of negative discussion > about Nikonscan - I really cannot see why people bag it > so much. My main beef with Nikonscan is the jaggies. Makes it useless for me. Aside from that problem, Nikonscan works very well, and has a far more usable interface for meaningfully adjusting scans than Vuescan. But we've already established that Vuescan's intention is to capture the most data possible and deliver it to Photoshop where it can be edited. > I get predictable output and generally excellent > colour 98% of the time from negs. I don't do much > slides, but they were fine too. When I was using Nikonscan, I did get quite good colour, with the exception of a roll of whale-watching photos. Nikonscan wanted to change the colour of the photos to something virtually B&W because of the dominance of blue in the images. In fairness, the default settings of Vuescan did the same thing (white balance) but the "neutral" settings did not. > I certainly get better results colour-wise than I > could ever get out of Vuescan, and VS was *much* slower. Odd. Vuescan is significantly *faster* on my computer, especially compard to Nikonscan with ICE. I suspect you are also getting significantly better results from Nikonscan since the LS2000 gives you access to the high bit options. The LS30 is restricted to 8 bits in Nikonscan and you have to be careful with adjustments to avoid posterisation. > I do agree that the jaggies is a real problem, and have not > been impressed by the results of my email discussion with > Nikon USA about this. I am about to send mine back for > "repair" re jaggies - I have little hope but will > report how it is dealt with. (remember this is in Australia). I'll be intrigued if Maxwell Optics manage to cure the vibration that causes the jaggies. As far as I can see it's a design fault caused by a combniation of hardware and software behaviour. > IME NIkonscan default auto settings cut off too much at the > high end (and maybe the shadows end too), so I use the > option - Scanner extras / prescan mode / low contrast neutral. If I ever get to use Nikonscan jaggy-free, I'll try this. :) Rob (about to travel south to Canberra) Rob Geraghty [EMAIL PROTECTED] http://wordweb.com
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
"Dave King" <[EMAIL PROTECTED]> wrote: > Do you mean jaggies are all through the image, or along the edges? The jaggies are through the entire image but are most noticeable on high contrast edges within the image. By "edge" I presume you mean the outer boundary of the entire image. The jaggies are regular slippage of the scan lines in a sawtooth pattern. They are apparently caused by the scanner mechanism moving at a particular speed - one which happens most often when using Nikonscan that scans in 64K blocks. Ed has coded Vuescan to scan line by line with a delay to prevent the vibration that results in the jaggies. If you're lucky enough to have an LS2000 you could probably also eliminate the problem by using the multiscan option since it causes the scanner to scan each line multiple times, slowing the mechanism down. The LS30 doesn't support single pass multiscanning. Rob
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
> "Dave King" <[EMAIL PROTECTED]> wrote: > > Do you mean jaggies are all through the image, or along the edges? > > The jaggies are through the entire image but are most noticeable on high > contrast edges within the image. By "edge" I presume you mean the outer > boundary of the entire image. The jaggies are regular slippage of the > scan lines in a sawtooth pattern. They are apparently caused by the > scanner mechanism moving at a particular speed - one which happens > most often when using Nikonscan that scans in 64K blocks. Ed has > coded Vuescan to scan line by line with a delay to prevent the > vibration that results in the jaggies. > > If you're lucky enough to have an LS2000 you could probably also > eliminate the problem by using the multiscan option since it causes > the scanner to scan each line multiple times, slowing the mechanism > down. The LS30 doesn't support single pass multiscanning. > > Rob I asked because I wasn't sure what you referred to. The term "jaggies" is usually used to mean "pixelization", which of course is a problem of too low resolution for a given output size, and it occurs in all digital images at some point. Nikonscan has two artifacts I can see, I believe both related to ICE. One is the "serrated edge" effect caused by scanning film in mounts, and it only occurs at the edges of the frame. The other I'll call "shark's tooth", and it looks like tiny spikes at regular intervals on high contrast edges. I think it only occurs in one direction, going from memory. To be clear what we're talking about, do you mean the tiny spikes? I'm asking for further clarification because from your discription, "regular slippage of the scan lines in a sawtooth pattern through the entire image", I'm not sure my particular scanner suffers from this problem. Dave
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
On Tue, 10 Apr 2001 09:04:19 -0400 Dave King ([EMAIL PROTECTED]) wrote: > The other I'll call > "shark's tooth", and it looks like tiny spikes at regular intervals on > high contrast edges. It's a regular, stepped displacement (on the y axis of a landscape scan) of pixels which repeats every 4-5 pixels. It is most visible on high contrast edges, but occurs throughout the image. You can see it in the full res LS30 scan at my site eg the boundary of her chin against the black background. ICE was not used for this scan. As you say, this is not related to the normal 'jaggie' phenomenon which arises through aliasing. Most, if not all, LS30's seem to do it occasionally or always, with Nikonscan. As the scanner ages it tends to worsen. I don't know if v3 helps, VS certainly does. - Regards Tony Sleep http://www.halftone.co.uk - Online portfolio & exhibit; + film scanner info & comparisons
Re: aliasing was Re: filmscanners: Review of the Nikon CoolScan 4000
> > The other I'll call > > "shark's tooth", and it looks like tiny spikes at regular intervals on > > high contrast edges. > > It's a regular, stepped displacement (on the y axis of a landscape scan) > of pixels which repeats every 4-5 pixels. It is most visible on high > contrast edges, but occurs throughout the image. You can see it in the > full res LS30 scan at my site eg the boundary of her chin against the > black background. ICE was not used for this scan. As you say, this is not > related to the normal 'jaggie' phenomenon which arises through aliasing. > Most, if not all, LS30's seem to do it occasionally or always, with > Nikonscan. As the scanner ages it tends to worsen. I don't know if v3 > helps, VS certainly does. Thanks, exact answers are reassuring even when they contain bad news. My scanner doesn't have very high mileage yet, and I'm thinking of tradin her in on the 8000 anyway. Further enticement. Are you thinking of mounting the effort to get your hands on one for review? (I hope I hope) Dave
