Extra-mural job opportunities

Image Reports’ February issue included a story on the giant mural at the Manchester Chill Factore centre. This contained a passing reference to an issue of gigabyte-sized imaging that may become more and more important in the next year or so. To save you digging out your copy, here’s the story in brief. The Image Group in Manchester printed and installed a 3,500 m2 mural on a series of vinyl panels to cover the entire wall area of the giant Chill Factore, an indoor winter sport/leisure centre. The finished printed walls measure about 350 m long and 19 m high.

The Image group has one of the new EFI VUTEk QS2000 2.0 m wide flatbed UV printers, which can output relatively high resolutions of 1,080 dpi at speeds that make it realistic to use in real-world production if the job demands it.

“The image started life as an 85 MB original picture of the Caucasus mountains,” said Dave Brunt, operations director at The Image Group. “Our pre-press team used specialist interpolation software and many careful Photoshop hours to increase the file size to 1.5 GB at 300 dpi before printing. The finished print is about 350 m x 19 m.”

Wide and super-wide format printers like this are increasingly able to output photo-realistic imaging at resolutions intended for close-up viewing. This may never be a huge issue because it’s always going to be quicker and more economical to print at lower resolutions and use less ink for billboards and posters that only need to be viewed from a distance.

However, as EFI VUTEk’s Kevin Currier wrote in IR May 07, digitally printed wall coverings are growing in importance. These will generally be viewed closer-in than digital signage, so ideally they’ll need high res originals.

In IR November we wrote about the new ‘blanc wall’ digital wall covering method developed by blanc canvas in Leeds. It’s a combination of scanning, plus a self-adhesive media and digital printing. The blanc wall service can scan paintings and fine art using a very high resolution rostrum scanner called iScan3D to capture both image and texture detail from conventionally-sized originals, and then to enlarge them to wall-spanning size.

Paintings and the like tend to be relatively large, but on the whole don’t contain ultra-fine detail, so they enlarge successfully. Photographs are different, whether digital or on film. There are some high resolution models, but they can scarcely function outside a studio. Professional digital cameras that you can feasibly use outside tend to have between 10 and 20 million pixels at present. If you blow these up to wall size you’ll see huge pixels. There are resolution enhancement programs to reduce this effect (we reviewed a bunch of them in IR June 07, such as OnOne GenuineFractals Print Pro, FixerLabs SizeFixer and BenVista PhotoZoom Pro, as well as Photoshop’s built-in resizing), but essentially they’re only hiding the pixilation without adding ‘real’ detail.

Film isn’t dead of course, and unlike digital cameras it’s analogue, so you can digitise it at any resolution you want. However it still has problems when blown up to large scale, because the limitation on capture resolution is the silver halide grain or dye clouds which starts to become visible if you wind up the scale factor.

Andrew Ainge at blanc canvas says that his service avoids the grain problem by printing photographs at A4 size on an Epson inkjet and then scanning that. However, this only hides the issue and doesn’t capture any more detail.

So, having listed all the problems, is there a solution? Can you capture very high detail photographs for very large prints? You bet. It’s fiddly, but certainly possible, with a technique that’s as old as photography itself.

You can take a lot of overlapping small photographs and build them up to cover the whole of a large scene, capturing both high detail and large areas. Specialist software can stitch and blend the images together seamlessly to form a single, very high resolution image file. This can then be enlarged and printed with fine detail.

Most of the software on the market is intended to create panoramas, covering wide angles all the way up to a full 360 degrees if you want. Some of it can capture up-down as well as left-right, to make an image where the viewer appears to sit inside the centre of a spherical image.

It’s always been possible to join conventional photographs together in this way, but it took a lot of darkroom skill. From the late 1990s on, Apple and other companies produced authoring software that allowed images to be stitched together seamlessly interactive viewing. The main purpose was to produce low res images for interactive ‘walkthrough’ tours of museums, houses and the like, but they’d usually also allow high resolution images to be stitched and then output as a printable flat file.

The market developed over the next decade and now there’s a fairly wide choice of software. Canon, Olympus and some others build overlapping aiming software into consumer digital cameras, and bundle stitching software to join it on a computer.

French developer RealViz is market leader in Europe, with its Stitcher 5.5 software. This is a smart package that makes it relatively easy to perform the ticklish task of aligning multiple overlapping images and blending them seamlessly. Removing distortion caused by lenses. Its resolution is essentially only limited by the power and memory of your computer, so it can assemble very high resolution files indeed.

Adobe’s latest Photoshop CS3 includes Photomerge, an incredibly smart feature that can intelligently blend quite badly mismatched overlapping images. It’s easier to use than RealViz Stitcher and it’s effectively ‘free’ with Photoshop, but it doesn’t have Stitcher’s sophisticating manual editing tools for correcting mismatches, and unlike Stitcher it can’t create 360 degree seamless imager, or output QTVR or similar interactive format. However for print, it’s great.

The really complicated part of multi-image panoramas is the capture stage. It’s easy to do a rough and ready set by simply hand-holding the camera and shooting an overlapping series. However, there’s a good chance that they’ll never quite join perfectly, so you’ll get ghostly double images around prominent features like door and window frames. The way to do it properly is to use a specialised panorama head on a tripod, which lets you rotate the camera around the nodal point of the lens (think of it as the optical centre where all the light rays cross over). Align it right and stitching is a doddle. However the heads are expensive (£100 for a cheap one such as Panosaurus, £400+ for a good one such as Manfrotto or Agno’s).

The concept of getting ultra-high resolution photographs is being promoted by a bunch of enthusiasts who are posting their results on the Gigapan website (www.gigapan.org). It uses live zooming/streaming software based on Google Earth (the mapping/satellite system that lets you zoom from the whole earth down to cars in your street). The expanding gallery of gigapan images include some impressive examples, such as a view of the Golden Temple at Amritsar that shows you the whole scene, but then you can zoom into tiny details on the walls or birds perched on the roofs.

Gigapan is a project of Carnegie Mellon university, which is working with a developer, Charmed Labs, to create an affordable automated robotic tripod head that will allow these huge images to be captured automatically even with cheap consumer cameras, by shooting very many overlapping pics with their lenses zoomed-in for maximum detail, and then stitching in software. The robot head is about to go into Beta and it may only cost around £200.

There is one fly in the ointment if you want to use high res panoramas for murals or vehicle sides. It’s called perspective parallax, which is the way objects in an image become smaller the further they are from the camera. With conventional camera lenses you don’t notice this much, but if you take a lot of shots with a camera pivoting on the spot, then you get the same apparently distorted effects as an ultra-wide angle fisheye lens. It’s more noticeable the closer the subjects are to the camera, and if they contain straight lines. Thus a row of buildings shot from the other side of the street will appear to curve away from the viewer at either side, with a dramatically curved pavement. If you project or print the image onto a concave curved wall, then it will appear natural, but on a flat wall it looks weird.

Andre Ainge at blanc canvas says that this is stopping him using true panorama images for his wraparound photographic murals – al though this can cover all four walls, each wall is flat, so the image appears to curve away. Instead it’s necessary to take conventional flat-field pics and fudge it a bit to get the corners to work.

One way round this is to photograph the scene from a long way away, when the effect is less obvious, but this isn’t always possible. Another method is to physically move the camera side to side along the scene, on rails if possible, and take a series of overlapping pics – here the camera-to-subject distance doesn’t change, so you get a flat image.

There’s a software fix too. Some programs (including Photoshop Photomerge but not RealViz Stitcher) give the option of distorting the image to remove the parallax, by progressively enlarging the image away from the centre. The result on wide aspect pic, is a sort of bow tie shape, which can them be cropped back to oblong with the lines looking natural. Enlargement means there will be less detail at the edges than the centre, but if you start with a very high res image this may not matter too much.

Super-resolution printing and super-resolution photography. Could they be the future of wallpaper?

Back to top