Making a giant print from an 11MP image
Making a giant print from an 11MP image
The editing process for a very large display print – high quality resizing
At a recent photography show, Innova were showing some of their products, covering the range of media they supply.
Amongst all their papers was the JetMaster Photo Panel image display system.
Keith looked at JetMaster a while ago, as a simple and economical method for displaying canvas ‘box’ prints.
Before the show we were asked if one of Keith’s images could be used for making a very large print.
Before the show we were asked if one of Keith’s images could be used for making a very large print. It would be created by tiling together a number of smaller prints, as you can see in the picture to the right, taken at the actual show.
The requirement was for an image that would print at 3 metres by 2 metres. The actual print was cropped a bit to fit the space, but it’s still pretty big.
The image Innova wanted was one taken of a sunset at Bandon, Oregon, overlooking ‘Face Rock’.
The image was taken with Keith’s old Canon EOS 1Ds camera, which although ‘full-frame’ is actually only 11 megapixels.
The article below, shows the process involved in creating a file that could be printed this large, without showing any obvious artefacts from the enlargement.
This doesn’t mean that any 11MP image will work – you do need a good lens, good camera and some experience of what will work at a large size, and how we perceive print detail…
Update Note: This 2014 article still (2019) represents my overall approach to making very large prints, However I’m now likely to try Gigapixel AI as my tool for the actual resizing. This reacts to pre-existing sharpening and noise in different ways, so a bit of experimentation is again in order. As well as the detail in the review, see a short article I wrote about using AI Gigapixel to enlarge another image from my 1Ds : Printing low megapixel images at large size
The photograph I’m going to work on is from a collection I took one evening, whilst staying in Bandon, Oregon, in October 2007.
I was using my Canon 1Ds with its 35mm full frame sensor. This was shortly before I got my current camera, the 21MP 1Ds Mk 3 (used up until the 5Ds arrived, since for my day to day work, the Canon 1D X offers no useful advantage [more details])
The weather that evening was changing very quickly, with a big thunderstorm going over at one point. You’ll see that I took quite a few pictures, trying to capture aspects of how the scene felt to me.
I’ve written up a bit more about my general approach to taking photos for prints in an article:
This current article is much more about the technical issues involved in enlarging an image, without undue artefacts, and making a print that looks good. Whilst I go through all the steps I ‘ve taken, do regard this article as more of a guide to the various stages you need to consider, not an absolute recipe.
The shots are all hand held at ISO 100. They are all fairly dark, since I’ve exposed so as not to clip too many highlights (I wanted to keep detail in the brightest parts of the sky and clouds).
Opening the image in ACR in Photoshop lets me have a look at potential lens corrections. I’ve used my EF24-70 2.8L at F/8 so corrections are minimal (you might notice that the corners are a bit lighter in the corrected image.
Of the three adjustments available, I generally find chromatic aberration (CA) the only one I regularly pick. If your lens has much CA, then it’s worth zooming in to see the corrections and fine tuning them. Just two pixels worth of CA can really show up at the sort of enlargement size I’m considering here.
On the subject of lenses, a good lens makes for a better print only if all the stages in your workflow are optimised towards your ultimate goal, the print. I’ve seen many a discussion about how much better a particular prime lens is, that show no evidence of it making any tangible differences to actual prints made by people. Of course, I’m prone to using cameras hand held, so what do I know about ‘real’ sharpness)
One area that needs a lot of care if you are going to create a large print, is the application of sharpening in your RAW image converter.
In normal use, this sharpening is something I’d include in my processing, it normally gets lost in the detail of any print, and for images for web use, then you are going to be resizing downwards and using more sharpening anyway.
The type and amount of sharpening at ths stage does depend on your camera sensor, to some extent, so it will be different for the 36MP sensor in a Nikon D800E, to my old 11MP 1Ds. However, at the scale of enlargement here, the differences between resolutions are perhaps not as great as one might first think (the lack of AA filter in the D800E wil help too).
For this big enlargement I don’t want any ‘automatic’ sharpening applied (that includes playing around with the various sliders in the inset (from the main panel) above.
It’s at this point I decide to use DxO Optics pro V9.1 for RAW processing. I know that it handles my 1Ds files very well, has good shadow detail and very good noise reduction (I don’t need the full ‘Prime’ noise reduction setting for this ISO 100 image).
There are several areas of the image, that as a print, will give the feeling I’m after, in viewers, that there is more detail in this image than the basic 11MP would suggest.
The sky and cloud have smooth areas of colour which need no detail sharpening at all.
The clouds on the horizon have highlight details that are only a few pixels across – these will very easily show sharpening/resizing artefacts if not handled with care.
The waves have lots of low contrast detail which if subtly enhanced will give the impression (against the smooth sky areas) of having a lot of fine detail.
I do have to turn off the ‘Lens softness’ adjustment in DxO Optics Pro. This can work well with many images, but introduces too many artefacts in fine detail when enlarging.
At 300%, the smoothness I’m after at edges is visible. This looks soft at this stage, but will make a big difference when enlarging.
I’m going to use one of my favourite detail sharpening plugins (Focus Magic – Review) to add a small amount of sharpening to parts of the image.
The sharpening is being applied to a duplicate layer from the background.
A two pixel radius shows some rather nasty edges in the sky highlights.
Backing the amount of sharpening down to 50% makes the results less harsh.
I’m showing the sky detail here, since it shows up effects of over sharpening quite clearly.
Keeping the detail very fine in those highlights is going to have a big effect at final print size
I’d note that you could produce several different RAW conversions of the same RAW file and then blend parts together, depending on detail.
However in this image, duplicating the background (as a new layer), sharpening it, and masking the two together will suffice.
Detail in the waves is much more amenable to this sort of sharpening.
Don’t be tempted to overdo any sharpening at this point, the idea is to end up with a print where the range of sharpness looks realistic.
Adding a layer mask to this sharpened layer allows you to adjust where the sharpening is visible.
The white parts of the mask (rightmost thumbnail) are where the sharpened version shows up.
I’ve applied the masking with a graduated fill (it’s a fairly abrupt edge, but still not a sharp edge).
By making the mask temporarily visible (as red in this instance), you can see where sharpening is masked off from just below the horizon upwards.
Note that I’ve not altered the mask to apply sharpening to the top of Face Rock itself, since high contrast edges like this are where I’m really trying to avoid artefacts.
Do remember that what works best is very dependent on image content, so take what I’ve done here as a guide to the principles, not a hard and fast recipe to follow.
We’ve not resized yet, so don’t be too concerned about the image being a bit soft. Looking at images at 100% and 300% magnification takes some experience – don’t get too hung up on what it looks like (yet ;-).
There are many ways of doing similar things in Photoshop, so if you find another sequence of events (or choice of sharpening tool) works better for you, then good. Remember that it is the print on the wall that counts, not your level of expertise in Photoshop ;-)
There are so many forum threads about the best way of resizing an image that you could very easily come to the conclusion that special software is the way to go.
Well, I’m of the belief that for many images a simple Bicubic resampling is fine, since the differences between other methods and tools are usually lost in inattention to capture sharpening and final print sharpening.
It’s the long arguments over 100% crops of fine detail, enlarged by differing means, that convince many that continually looking at images at 100% (aka pixel peeping) is a bad thing.
The enlargement is so much that a wayward few bright/dark pixels (cloud edges, I’m looking at you) could easily become large enough to be obviously visible in the final print. If you have another preferred way of enlarging, then do try it, but be careful to note that it is not introducing any sharpening into the image in areas you don’t want it.
I’ve kept the print resolution down at 240ppi since we are looking at a huge print, and 300ppi would make the file so huge it would need to be saved as a .psb file (it would also go over the 30k pixel limit for JPEG files, if I wanted to send a more reasonably sized 8 bit JPEG version for printing).
With a linear magnification of around 7x, you can see how the tiny alterations in sharpening at the RAW and post conversion stage have such an important impact on the final print.
The linear magnification, if I was using a 36MP image, would still be ~4.6x, so don’t assume that a lot of my care in making this print is only needed because I’m using an image from an ‘old’ camera. The area of the image above is where Sony’s inability to store lossless sensor data in their RAW files would be painfully apparent, if you were using one of their cameras such as the A7r (see Thom Hogan’s review for more technical details of this problem – fortunately addressed through firmware updates for some models in 2015))
2018 Observation: I now use a 50MP Canon 5Ds so making such a large print is a bit easier, although the same principles apply. A new re-sizing tool [AI-Gigapixel] offers some interesting features for the big print maker, but still needs some care in the production of the images you feed into it for optimal results (see the review).
Note that I’ve left dust removal until after enlargement. Imperfect removal at original size can leave visible remnants when scaled up.
One way I make dust more visible is with a temporary curve adjustment layer.
Mouse over the image below to see the effect.
A steep curve can make low contrast dust spots much easier to find.
After removing dust spots, you can delete this adjustment layer.
As an aside, see how the curve has increased the visibility of noise in the temporary view.
This is also the time that I’m addressing the slight tilt of the horizon in the image.
This shot was hand held, I didn’t have a tripod with me. I generally dislike their use for landscape work (unless it’s dark) even though I use them for a large proportion of my architectural and commercial work.
After the rotation, I do need to crop the image slightly, but at 0.35 degree it isn’t much (if it was a problem, then I’d likely get cloning.
Do you see those birds in the picture above? Only a few pixels, but my care in the pre-resizing stages has left them clear, and without artefacts. A few tiny details like this, in a very big print can give the viewer the impression that there is a lot more detail in the view than there really is.
At this point I decide to add a bit of local contrast enhancement with Nik Viveza. Just a bit to bring up contrast for features like those crepuscular rays from the setting sun. This image is for a print, so you can push this aspect a bit, since prints have a much lower range of contrast than you see here. I’ve looked at other plugins that offer this sort of contrast enhancement, such as MacPhun Intensify Pro , Topaz Clarity and Topaz Detail – just be careful not to do too much.
Even with the light touch from Viveza, I felt that the effect was not needed at the top of the image – this view is of the mask I’ve added (shown in red) where I’ve ‘painted out’ the adjustment.
Whatever tools you choose, it’s worth using them on a duplicate layer. I find this allows a lot more subtle application of the effect. Resist the temptation to push adjustments too far – there were some quite ghastly prints on many of the stands at the show, showing that ‘turning all the dials up to eleven’ is still more fashionable than I’d hoped to see ;-)
Since we’re printing onto canvas, I’m also bumping up the vibrance to offset some of its tendency to ‘flatten’ images.
I’ve now got a huge file (~1.3 Gigabytes) at 16 bit, and in the Adobe98 colour space.
Print sharpening is where I try and compensate for some of the effects of printing an image. This is part of where people are disappointed with what comes out of their printer.
I’m working on a calibrated and profiled monitor, at a relatively low brightness (one of the principal problems when people ask “Why are my prints too dark“)
I need to apply a level of sharpening to the image, to compensate for the printer and media I’m using.
I’m using Nik Sharpener Pro 3, which can give very good results.
The effect on the water is far better, after I’ve reduced the sharpening strength (mouse over to see).
Remember that this is now an image for printing, so a bit more noise is not an issue.
The sharpening does however find a few dust spots I missed…
Although I’m happy with the level of sharpening applied to the whole image, there are a few area I choose to mask out.
I don’t expect you’d be surprised to see where… (shown in red)
I don’t need too much precision in this – you can see where I’ve painted out the sharpening along the top of the rock
Here’s part of the final image sent off to Innova.
I’ll admit some trepidation when I went to their stand. I do all my own printing, so I know what happens right the way through the process.
My huge file crashed their tiling software, so a few workarounds were needed, but the print(s) were OK.
The image has lost a bit of punch on the unvarnished media, but detail looks good.
I’ve sharpened this image knowing that some people would be looking at it quite closely. If I knew it was going to be in a location where no-one ever got any closer than say 10 feet, I could raise the overall sharpening amount in the final print sharpening. For prints this size, knowing the viewing conditions are important.
A closer view shows where I feel all that care with sharpening (and not sharpening) has worked well.
If I was making the print for myself, there are a number of additional bits of sharpening and masking that I might have introduced in the editing, but this sort of work needs test prints of small parts of the image to see how differing adjustments affect the final print. I do actually produce large prints myself, but more usually from 21MP Canon 1Ds Mk3 files or multi shot high res images produced using software such as Autopano Giga.
I couldn’t resist asking a few bystanders what they thought, and the look on people’s faces when I told them it was from an 11MP camera made it all the more worthwhile. There is more info about JetMaster at their web site
I do hope the general principles I’ve outlined in this article are of help to people looking to create large prints – please let me know if it helps, or there are areas of my choices that are not clear?
Never miss a new article or review - Sign up for our Newsletter (2-4 a month max.)
Enjoyed this article?
More print related information
For information about other printers, paper reviews and profiling (colour management) see the Printing section of the main Articles and Reviews page, or use the search box at the top of any page. There are also specific index pages for any articles connected with the following topics:
- Digital Black and White
- Tutorials and 'How to' articles
- Colour Management
- Printer test images
- Why do your prints look wrong?
More of Keith's articles/reviews (Google's picks to match this page)