Nikon D70: Interpolation query for A1 poster

[alistaircotton]

I have an A1 poster print job to run for a client and it looks like I

will need to generate/work with a 300dpi image at 837mm x 590mm. I

was hoping that my new Nikon D70 would do the trick but suspected

that around A3 would be a tall ask from the camera, let alone looking

at A1.

 

So I've run some bicubic interpolation tests with Photoshop and am

highly impressed with the results. Arguably this poster picture

should preferably be shot with a medium format camera, but a 35mm

tranny should suffice. So I've scanned one of my older 100 ISO

trannies with a CanoScan 4200f. I know it's not a drum scanner or

anything, but in theory should have sufficient grunt to produce a

better quality image than an interpolated 6 megapixel Nikon image.

 

Not!

 

I scanned the tranny in a variety of ways, including using sharpening

masks and the like at more or less 837mm x 590mm. I then took a Nikon

RAW image into Photoshop and enlarged it with a standard bicubic

transformation to the same size.

 

I then zoomed into both of the resulting images at 100%. On screen

the interpolated Nikon image appears to be at least a little bit

cleaner and sharper than the scanned tranny.

 

Then for the crux - I sharpend the Nikon image first with the bundled

Picture Project software before importing to Photoshop and then

running the bicubic transformation.

 

I kid you not, zooming into the image at 100% produces a near-perfect

photograph on screen.

 

So here is the question:

 

On screen, the sharpened and interpolated Nikon image beats the

panties off the scanned tranny. Should I even bother about spending

money on slide film and a drum scan to A1 when it looks like my Nikon

D70 is more than up to the task?

 

On second thoughts, is there something I'm overlooking with regard to

this?

[._kaa]

Two observations.

 

First, you realize you're comparing your D70 to your scanner, right? Take the same tranny and make a proper drum scan -- I would expect the results to be quite different :-)

 

Second, how enlargeable the digital image is depends very much on the image. Say, a landscape will lots of fine details won't enlarge too well. But take sunset clouds with no sharp edge anywhere, and you can enlarge a digital image to almost any size imaginable.

[beauh44]

I certainly won't argue that a drum scan will get much better results than most home-made scans. But man, they're expensive! Add to that film and development costs, and the bucks can add up fast. Also chances are, even with drum-scanning, you'll likely still run up against film grain and I suspect that your Nikon, on its lowest ISO setting will be almost noise-less and will thus interpolate very well. I've got a bunch of film cameras but don't use them much any more. Good luck!

[awindsor]

A drum scan will extract slightly more shadow detail than a dedicated 35mm film scanner or the best of the new flatbeds. This was discussed recently at photo-i in the Epson scanner forum. Quentin compared his Howtek D4000 drum scaner against a Epson Perfection 4990 flatbed scanner. The drum scanner extracted more shadow detail but the difference is pretty subtle. He rated the 4990 equal to his old Imacon. Realistically unless you need the extra shadow detail you will do as well with a consumer scanner. These comparisons were willout wet mounting on the flatbed.

 

I think 35mm is inferior for large enlargements unless you go with a lot of noise reduction software since even with a slide you get "grain" which becomes objectionable. The digital files enlarge easier though there is a "emptiness" resulting from a lack of fine detail. Fine for a portrait, or a product shot, but I find it objectionable in a landscape or the like.

[christian_mozetic]

I drum-scanned 6x6 Kodak E100VS to 1x1 meters for a job and was very surprised by the amount of grain the scan picked up. Of course, I'd never printed a 1mx1m poster before so I had no reference for comparison. I compared the scans to images I had taken on that same job with my 5Mp Sony f717 and of course there was much more detail on the MF scans.

Below you can see what the sky looked like on the scans at 100%. It was scanned at about 5000dpi.<div></div>

[mark u]

If you look at a projection of your slide to the required size, I'd lay odds (providing you are using a fine grain film and a good projector and screen) that the image looks much better than the upsized scan - even viewed at a "pixel peeeping" few inches from the screen. The problem is that image processing software is mostly designed to handle regularly spaced pixels - whereas film has a random distribution of "pixels", which are then sampled on a regular grid by the scanner. Effectively, a projection lens is much better at image processing to produce a pleasing result. You should look carefully at your image processing technique to reduce the effects of grain, and also consider using better upsizing software - such as Qimage or Genuine Fractals to handle your scan. You may find that these programmes make a better job of upsizing from digital originals too.

[jorgen_udvang]

To make this even more confusing, different printers and printer software, will handle different kinds of "imperfections" in the photos very differently. An offset print based on a 300dpi file doesn't always give a better result than an inkjet print based on a 150dpi file when both are based on the same photo.

 

The best way to obtain an optimal result, is to experiment with the printing equipment that is going to be used. See what base resolution will give the best results on the final print.

 

If offset printing is a requirement, and you are in the borderlands of how much you can enlarge your photo, it is sometimes a good idea to make a file that is around 50% the size of what you will print, and do the final enlargement with the postscript file generated at the film/plate-shop. I'm not sure why this works, but it often does.

[jcuknz]

It is mentally stimulating to do all the exercises you are putting yourself through BUT .. how close are the viewers going to be to the final print? If the answer is not very close then the exercise is unneccessary since one of the parameters of resolution and depth of field is the viewing distance.

[alistaircotton]

Thanks for all your comments and suggestions with regard to my query. On seeing Christian's image sample I ran a quick test to see how a seriously enlarged RAW image would cope with a massive bicubic enlargement.

 

Flower 1 (lower quality) was sharpened to a medium setting in Nikon's Picture Project and then imported to PS. Once in PS I resized the image to a square (effectively losing pixels). I then enlarged the image to 1mx1m. This picutre is viewed and saved at 100% in PS.

 

Flower 2 (slightly higher quality) was sharpened in the same way, but to the maximum setting in Nikon's PP. It was imported to PS and enlarged directly to 1m wide (a happier solution for 35mm but not really a direct comparison).

 

I would not want stand right next to either of these images, but it may view okay from a distance and does not appear to compare too unfavourably with Christian's drum scan with 120 film.<div></div>

[alistaircotton]

This is flower 2<div></div>