Epson V750 vs Imacon 949

[timparkin]

I'm pretty sure, its from left to right:

Tristan 7500
Tim 4500
Ian Imacon
Tim V750
Joanna V750

yep...

[timparkin]

As the owner of the transparencies:
In terms of colour balance on the Wild Garlic:
1-Ian
2-Tim -Howtek
3 - Tim - Epson
4 - Tristan
5 - Joanna

Interms of shadow detail on the Wild Garlic:
1 - Tim Howtek (by about 5%)
2 - Ian
3 - Tristan
4 - Tim Epson
5 - Joanna

There are 2 other transparencies 1 of Appley Tower (the Castle) the other a close up of a fern. In terms of colour balance, my own are the most accurate, which reiterates the old adage that 'you' know your own images the best.

I'd obviously disagree - but then again this could be because we're using two different monitors and different light boxes. For my money, The Wild Garlic shots are a tough call on colour balance and I'd have to see the transparency again to judge but I'd say it's probably in between somewhere. The biggest difference appears to be in the sky where mine is more egg shell and Ian's is more cyan. Also, my greens tend to be more yellowy green whereas Ian's are purer green.

As for the castle picture, I would have to say that Ian's seem to have too much of a yellow bias in the warm stones and a bit of magenta in the skies that I can't remember seeing.

The fern has quite dramatic colour and brightness differences with ians being very cool and mine quite neutral.

These differences could well be down to lightbox colour temperature and screen temperature.

I'm hoping to get Joe to scan them and it would be good to send them to DJ too as he seems to have a very nice profiling setup.

However, I'm interested in how much work you did to match the transparency to the scan? Obviously if your profiling is up to scratch you should need no tweaking.. Any tweaking may introduce errors from the lightbox and monitor? I'm going to include my raw, untweaked scan in when I make the comparison also - do you still have your original ian?

Probably a good idea to get some external judges. I'll get a few people to have a look at them as a blind judging and see what they say.

[Patrick Dixon]

Just get blind people to look at them ...

[timparkin]

Just get blind people to look at them ...

It's all about the feeling after all ...

[Charles Twist]

A little bit late, but...
I use the same targets as Joanna and a V700, and I am fairly confident that the noise in her scans is not caused by the profiling target during post-scan processing. I suspect it's a limitation of the equipment, probably made worse by some sharpening. On Tim's recommendation, I have tried scanning at higher resolutions, then down-sampling. This gives very clean shadows which can be pushed a lot without ill effects. I am not sure whether this is tantamount to averaging out the noise or increasing the apparent sensitivity of the sensor, but who cares?
The downsides of scanning at 3 or 4 times the final / usual dpi are (i) the longer processing times - indeed, my PC does not have enough memory to handle a full tranny at 4800 dpi in 16 bit and (ii) the ensuing impossibility of printing greater than A3 from a tranny with dense shadows. But then, the point made at the beginning of this long thread, is that the V700 is perfectly adequate for A3. So no surprises there.
From an aesthetic point of view, there is a risk of over-egging the shadows then ending up with strongly contrasting blotches of colour and black. Caveat emptor. From the results, I don't get the feeling that the drum scanners have greater Dmax, ie a darker cut-off point, but I would like to hope so. This would counter the blotch effect to some degree. I also recall Tim experimenting with a modified target profile so that the blacks weren't so black. Tim, did you use that here?
Final consideration regarding the colour of shadows and the effects of memory and emotion on perception: I recommend very strongly a look at landscape paintings made during the late impressionist period (eg Monet and even Van Gogh). It's quite striking how they render the shadows and highlights. In some cases, the results are akin to HDR, with very little tonal variation, but an extraordinary sensitivity to hue - very often blue shadows and yellow highlights. In the absence of spectrometers or colour photography, their pictures demonstrate remarkable perception abilities. Coming back to film and targets, there is so much minor variation with natural light, that I doubt small variations caused by profiling with different targets would be picked up, even by the original photographer. Of course, if you place the two prints side by side, you can say this is more cyan or that is more yellow, but looking at the print on its own (absolute not relative viewing), would you turn either down as not worthy?
Regards,
Charles

[joolsb]

Charles,

Received wisdom has it that scanning at 4800dpi on an Epson is a largely pointless exercise as the scanner starts to interpolate above (I think) 2400dpi(possibly lower). So scanning at 4800dpi and then downsampling is effectively a double-interpolation. Now, if Tim says this leads to an improvement then I'm sure he has the results to back up his claims but, all the same, I'm a tad sceptical.

[Charles Twist]

Hello Julian,
Well it's not entirely pointless going on the evidence.
Let's say we have a scan which gives one pixel which is not v. dark but green, ie noise. Now let's say we increase the dpi so we have 4 pixels rather than the 1. Not all 4 will be green; in fact, it stands to reason that more than half are not green going on the results. I am not sure where the Epson gets the extra info from, but the software doesn't simply replicate the one pixel or make it larger. Which again points to the process being pointful. So when you down-sample in Photoshop or other - this is a critical step - you lose some of the green pixels you would have had, if you had only 1 pixel not 4. This is obvious whatever your starting dpi - a smaller print has less noise than a larger one, in absolute terms (but probably not in relative terms as there is also less information).
Regards,
Charles

[Marizu]

Hi Charles. Do you suspect that the scanner effectively "over samples" (or samples at a finer precision than its mechanical precision would allow) at 4800dpi but that only some of these samples are contaminated by green noise which is then normalised away upon downsampling to the 2400dpi limit of the scanner's detail resolution?

[Charles Twist]

I really don't know what's going on. I didn't think of trying 4800dpi until Tim mentioned it. Maybe he has a reference?
Regards,
Charles

[Michael J Finch]

Not strictly to do with this thread but might be worth a peep at Silverfast's take on resolution: http://www.silverfast.com/show/resoluti ... et/en.html
Does anyone know if there are other targets out there that don't cost the thick end of £75?

[timparkin]

Hi Charles. Do you suspect that the scanner effectively "over samples" (or samples at a finer precision than its mechanical precision would allow) at 4800dpi but that only some of these samples are contaminated by green noise which is then normalised away upon downsampling to the 2400dpi limit of the scanner's detail resolution?

Hi Marizu,

The noise that you get is pretty random and every pixel that is scanned is a toss up as to what level of noise it has. However, if you over sample. Your noise will be at twice the frequency of the details in your scan e.g. your noise will be at 6400dpi or 4800dpi whereas the scanner will only scan real detail at just over 2000dpi. This means you can apply noise reduction at 1-2px radius and not affect the actual image detail (or film grain clumps) this means you get less noise and that you get is less obstructive.

Tim