Working with very, very large images

By Amos Willbond


Liebrand Case Study

This case study briefly describes the technique and steps that Ben Liebrand of Liebrand Images took to create a 10m X 3m graphic to advertise a Swedish Radio station on the side of an articulated truck.

The design brief stated that the image was required to be high resolution which, given the overall dimension, meant that using a conventional bitmap editing package would present a few problems with the composition of images. The main disadvantage with using a ‘Pixel Pushing’ image editor was speed vs. machine specification as these type of programs require available RAM to process all images. As the final output was calculated to be around 500Mbytes in size, the project file would require around 1.5Gbytes of memory to be able to process at an acceptable speed.

Using a vector based program would give the user comparable speed, but it was soon found that very few such programs existed with the capacity for compositing between multiple layers and accepting the necessary Photoshop compatible plug-ins.

Ben had heard about Satori and felt that this was a suitable time to give it a try. Liebrand Images subsequently generated the advertisement in a very short time, keeping productivity as high as possible.

Firstly, elements of the image were generated in a 3D package at high resolution for compositing within the final image. All the rendered 3D images were then converted to the native Satori file format for optimization within the program. This is especially necessary when using large image files or very complex canvasses containing many layers and can make the difference between the program being painfully slow to being lightning fast. The first guide layer, a low-resolution Photoshop image that had been the concept approved by the client, was loaded directly into Satori. This was now the canvas where everything would be generated, yet it was only 800x250 pixels in size…

After generating a new layer the speakers were placed into the image by dragging a rectangle in the correct area and simply assigning the pre-rendered image to it. This rectangle was then duplicated twice to add two more speakers. To get the speakers to fade into the background, the mask channel was selected (by clicking on the mask icon on the main menu) and airbrushing was applied. Further elements were added to the canvas of either 4000x4000 or 6000x4000 (source) pixels. The image was completed after 14 layers, still at the initial 800x250 pixel size, but containing enough image definition at source to generate the final, truck-sized image

Working with so many large images, essentially cutting and pasting 80Mbyte files onto a multiple layer project, you may expect the computer to start to slow down on about the eighth or ninth layer. "I completed the composition comprising of 45Mbyte and 80Mbyte images with the same kind of speed as if I was working with 1Mbyte images", said Liebrand, "when the composition was finished, I saved the canvas (which holds the formula for the composition) which was 5Mbytes. So apparently you could have loads of versions of one project without filling up the hard disk."

This saved project file could now be loaded up at any time to convert the file for output to almost any media.

Also it may be worth noting that anything applied to the original image can be edited when you load this project file back, It’s not a multiple step undo (something that seems to have generated a lot of hype in the past….), because it applies to EVERYTHING.

Importing the canvas file back into Satori, still at 800x250, it was time to render the output .tif file to finish the job.

The canvas properties were changed by entering new values into the dimensions dialogue boxes of 978x317.6 cm (or 385 inches by 125 inches at 50dpi, 19251x6251pixels in real money) and more airbrushing was added, interactively in real-time, with a brush size of 2000 pixels. Using the ‘Save Bitmap’ command, a filename was entered and the appropriate file format was selected. When the file was originally saved, Satori immediately reported that the page-file needed increasing to a whopping 960Mbytes. After setting this, Satori was able to render the output .tif file while the designers carried on with other projects, using the batch render facility meant that they could even work on the same machine that was rendering. The next morning the final image had been rendered and the resulting file stored to disk. Ben estimates that this took around 5 hours on a 266MHz Alpha. "This may appear to be time lost, but it was actually time gained, as I continued with the next project, right after hitting the render button". That, added to all the time saved while actually manipulating and compositing the image and the infinite undo, gives Satori users the chance to be more creative and free to make the mistakes necessary to be innovative, without sacrificing their workflow.

Reloading a composition like this takes a few seconds and it is possible to generate bitmaps of virtually any size, meaning that you are never short of resolution. Let’s face it, it’s happened to us all, the client originally wanted a graphic for use on a web-site, then liked it so much the company decided to make it into a poster. Where do you get the extra resolution? In Satori you just have to change the size of the canvas……