Post processing IQ180 files for resolution/camera test

[theguywitha645d]

A digital image where you see pixelation, jaggies and aliasing when you come close is *not* nice. There's no romance to digital artifacts. Say if your digital file lets you keep the print resolution at 200 ppi or higher there will (generally) be no problem. But say if you will make a *huge* print so you get down to 10 ppi or so it will become more important that the file enlarges well than that it has the highest resolution.

I can see you have not done too much large-format printing with digital files. The RIP and print drivers actually interpolate the file data so pixelation is avoided. There are no jaggies. It really is a myth that there is a limit to print size with digital files, just like folks used to claim there was a limit to film images depending on format size.

[hjulenissen]

I can see you have not done too much large-format printing with digital files. The RIP and print drivers actually interpolate the file data so pixelation is avoided. There are no jaggies. It really is a myth that there is a limit to print size with digital files, just like folks used to claim there was a limit to film images depending on format size.

I havent done too much large-format printing with digital files. But if the image sensor/raw converter is doing nasty tricks to increase "perceived sharpness", I believe that it will be hard to get the image in a form that is suitable for scaling?

In principle I agree with you, though: if the image is sampled according to Nyquist (or close enough), then theory predicts that it can be scaled to any pixel grid using a simple linear filter with no artifacts (i.e. the image would appear more blurry the closer the viewer moves towards it).

-h

[torger]

I can see you have not done too much large-format printing with digital files. The RIP and print drivers actually interpolate the file data so pixelation is avoided. There are no jaggies. It really is a myth that there is a limit to print size with digital files, just like folks used to claim there was a limit to film images depending on format size.

I have printed under-resolution pictures though, although not billboard size :-). I also want my normal prints to look well under the loupe :-), so I am a little familiar with the problem. Printer drivers usually do something similar to bicubic-soft, and it will work well if your original image does not have too high pixel-to-pixel contrast, or if your end ppi is high.

Attached an image showing an example, upscaled with bicubic and viewed at 200% for clarity. The left picture is an upscaled version of a pixel-sharp original file, as it would look without AA-filter and sharp lens (and no heavy diffraction), the right is also upscaled and is based on a higher resolution original file, but with AA-filter and fairly soft at the pixel level. On the left you can see where the original pixels where around high contrast edges, while the right will not show that regardless of how much you upsize it since the original file is soft enough.

Exactly how soft a file needs to be to make pixels 100% invisible with bicubic upscaling can probably be calculated and defined mathematically (some maximum contrast level between neighbouring pixels). I know many digital systems are too sharp, especially if the file is sharpened before upscaling.

For more examples of what artifacts I mean you can look at http://www.benvista.com/photozoompro/examples they show upscaling of initially downscaled images and thus corresponds quite well to a system with very sharp lens and no AA-filter (although even such a system does not get *that* sharp, and there's demosaicing). Anyway, bicubic upscales has very clear artifacts and visible pixels. Those examples also show spline upscaling which hides pixelation but I think gives too artificial look and false detail. The best is to start off with a file which is soft enough at the pixel level so it can be upscaled bicubically to any size.

[torger]

In principle I agree with you, though: if the image is sampled according to Nyquist (or close enough), then theory predicts that it can be scaled to any pixel grid using a simple linear filter with no artifacts (i.e. the image would appear more blurry the closer the viewer moves towards it).

In many digital systems, especially medium format (?) the resolving power of the lens etc is higher than of the sensor, so you can get aliasing moire etc. Some sensors have AA filters to reduce the resolving power to what the sensor can sample to avoid or minimize that problem. Generally, high res photographers dislike AA filters since they reduce contrast at pixel level, i e the percieved sharpness and detail becomes less when pixel-peeping. However, while such files are nice when printed at high PPIs, it seems to me that they do not enlarge well to billboard size if you want them to look pleasing up close (i e just soft rather than soft and pixelated).

Film does not have this problem.

[theguywitha645d]

All I can say is I do not have your experience. I just made 7 foot prints from my 645D which has no AA filter and there is no pixelation. I also print 12 foot panoramas using the same camera, still no problems. If I made a billboard, the image would continue to look fine--you view them from cars after all.

Naturally, film image if optically printed start suffering at huge enlargements. You can digitize them, but then you are back to pixels.

Still, there is no limit to print size to either a film or a digital image. One is not better than the other.

[torger]

I was trying to explain why some has the experience that film enlarges better (like "harlemshooter"), and why if you scan film that you should use much higher resolution than the actual detail film has captured, for huge prints you want to reproduce the grain too.

However, your mileage will vary. How aliased or not a digital image will be depends on many factors, sensor pixel pitch, lens sharpness, aperture, focus, subject etc.

Here's an example of a Canon 5D without AA-filter to the left and with to the right.

http://www.maxmax.com/images/Cameras/Canon%205D/Comparison_3.jpg

I'd say that the left one is too aliased to enlarge well to huge sizes, but if printed at say 300 ppi in a book it will probably look better (apart from the moire) and sharper. An old 5D has 8.04 um pixel pitch though, which is not a tough challenge for lenses to render very sharp. More recent sensor with smaller pixel pitch will have less of a problem due to that the lenses (and possibly diffraction) will soften the image some. IQ180 has 5.17 um pixel pitch, together with lenses it probably will not render images as sharp-per-pixel as this 5D example, but probably sharper than the 5D with AA filter.

[BillOConnor]

Can we talk about the realities of using an 8x10 view camera vs. the realities of using an MF camera with a digital back?
The 8x10 camera has such shallow depth of field, 325mm normal lens makes that a reality, that the photographer must choose his subject with great care. Swings and tilts only have limited effectiveness at 300mm and up, even at 150mm and up. The old shooters chose f64 because it leveled the focus field. The diffraction fuzzed the sharp areas so that they more closely matched the out of focus areas and the photographer relied on brute tone range to carry off the illusion that everything was mostly in focus.
Outdoors, wind was a brutal foe, a breeze not much less cruel. The film sagging in the holder helped a little because the curved piece of film more closely matched the curved field of focus of the lens, or not.
The equipment is brutally heavy, slow to set up, slow to use and completely impractical for magical times when the light is on the run.
Comparing the use of an 8x10 to the use of a MF camera with a digital back is a dilittante's folly. If you are comparing the ideal result in resolution from a scanned 8x10 transparency to the ideal result in resolution from an 80 megapixel back, you've achieved a result not even worth worrying about because the number of people shooting 8x10 sheet film will soon equal the number of people shooting 8x10 daguerreotypes.

[Stefan.Steib]

>>>>>the number of people shooting 8x10 sheet film will soon equal the number of people shooting 8x10 daguerreotypes<<<<<<

Maybe even less..... Nobody knows what will happen if Kodak stops making sheet films. With only Fuji left and those also cutting down the Sheet film program this is only a matter of time until
also the Japanese will pull the plug. Kodak has sold the pearl of his digital assets now (which is not clever as I think - it´s probably only a desperate move to win time ) :

http://uk.reuters.com/article/2011/11/08/eastmankodak-idUKN1E7A61RD20111108?feedType=RSS&feedName=technology-media-telco-SP

and they are further looking into cutting non profit departments. They may gain some more time now to solve their problems, but I am sure sheet films are another victim to
streamlining their portfolio.
And what will be left over after this - a printer department which is just one among many and not very successful (with losses )......... ?

I wish them luck, Kodak has brought the world a lot of great inventions and I have worked with their products for a long time, but I fear they need more than luck.
They need a new concept and film is not part of this.

Greetings from Munich

Stefan

[Ben Rubinstein]

hehe, this is only the dust on the tip of the iceberg of what the responses will be like once the test is published

Tim, wishing you all the very best and the strength to ignore the flak.

[buldoozer]

Some years ago the global thingking was that you need 50MP to beat 4.5X6 ,200MP to beat 4X5 and 800MP to beat 8X10 format.A lot of thingking mistakes has been made at the time.The overal botleneck whit analog wed photography has always been the grain.The picture was captured on the grain so fine grain and negative size were importend.This is whit digital clearly not the case or bottleneck.Whit digital the picture is not captured on the sensor bud on the compactflashcard.The sensor works only as some sort of a transformer transforming fotons into electrons.Digital has no grain so al the problems which come whit it are gone.Thats means that sensorsize has nothing to do whit picturequality anymore.It has only to do whit how large you want to print the picture.The same goes for pixels.As long you cannot see the pixels on the final print whit your nose close to it,the resolution of digital is actualy infinity high.Thats why we now see in 2012 that digital prints are completly clean and sharp whit infiny high resolution because there is no grain at oll.As long as you keep the ISO low enough,you whont see noise too .I have razor sharp clean prints  from my 5D2 at 24X36 inch whit out pixelasation at 158DPI.You dont need 350DPI on a print to get a sharp print.350DPI was based on wed photography.Now in the end of the analog versus digital discusion whe see that 18MP is enough to beat 4.5X6, 40MP is enough to beat 4X5 and finaly 80MP is enough to beat 8X10 analog  negative format.Even the sensorsizes are away smaller than the big negatives that were needed in analog photgraphy.Bud how far does it go?. Sony has come whit a 24MP APS C format camera.Greating to you all from the Netherlands.