Montreal in gigapixels

An idea that stayed with me

I discovered gigapixel images in a web experience created for Ubisoft around Assassin’s Creed Unity. From what I remember, the view looks out over Paris from high up on Notre-Dame. The scene depicts the French Revolution, and it’s huge. You can zoom into the crowd, look at the characters and find bits of the story hidden in the image.

I keep the idea in the back of my mind without thinking much about it. Ten years later, one summer day, it comes back to me: could I do this with my camera? I quickly look up how to do it online: take lots of photos while zoomed in, stitch them together, then split the result into layers, a bit like Google Maps.

The setup

I bring a Nikon D3200 I bought about ten years earlier, a NIKKOR lens set to 300 mm and a $30 tripod. It’s what I have on hand. Without a motorized panoramic head, I’m going to have to move the camera myself between shots.

The telephoto lens lets me isolate a small part of the city and photograph its details. With the tripod, I keep the camera steady while working my way across the view, row by row.

The shoot

The idea behind a gigapixel image is to stitch together these small parts of the scene photographed with a telephoto lens. I cover the view row by row, keeping roughly 30% overlap between photos so I can join them together. Exposure, white balance, focus and focal length stay fixed throughout the series. I save the photos in RAW.

At 300 mm, the slightest movement shows. Every time I move the camera, I wait for the tripod to settle before taking the next shot. Those pauses of a few seconds add up across the grid. The shoot takes about three and a half hours.

During that time, the light changes, shadows shift and things in the scene move. I also need enough overlap between neighbouring photos and between rows. That’s what keeps gaps from appearing in the final image.

A motorized panoramic head makes the movements easier and the shoot more consistent. My $30 tripod is enough to try the idea. It just takes more patience and attention.

Stitching the photos

Once the shoot is finished, I use stitching software to align and merge the photos. The overlapping areas provide the matching details needed to join them. Then I need to check the whole image, fix any joins that need attention and crop it.

I mainly check sharpness and differences in lighting. Anything that moved also needs attention: it can appear in more than one place or get cut off at a join.

After overlap and cropping, fewer pixels remain than the total from all the original photos. There are still several billion of them. My working file is around 4.5 GB in PSB format, the Photoshop format for very large documents.

From image to browser

To put the image on the web, I use libvips^[1]. Its dzsave command generates a resolution pyramid in Deep Zoom format.

This produces several versions of the image at different resolutions, each split into tiles. The browser only loads the tiles that match the area you’re looking at and the zoom level. You can explore the image without downloading the complete file.

For the viewer, I use OpenSeadragon^[2]. It handles panning, zooming and loading more detail as you zoom in.

These two open-source tools give me the technical foundation. I can then work on the interactions and content to attach to the image.

Zoom as a storytelling tool

What interests me most about this format is making content appear at different zoom levels.

You can attach text, sound, video or animation to an area of the image and reveal it at a specific scale. A character in a crowd, a scene behind a window, a clue in a reflection: each can become the starting point for an interaction.

Visitors choose where to look and in what order to discover the content. You can give them markers to guide their route, build a search for clues or leave them free to explore.

For a brand, this can be a way to present a destination, tell the story of an event or showcase a collection. Composing the image for the occasion also lets you bring several stories into the same space.

Those details are part of what makes the experience interesting. You need to think about what catches the eye, what appears as you get closer and what makes you want to keep looking elsewhere, while keeping the overall view in mind.

What I take away

I’m really happy with the result, especially with this gear. Moving the camera by hand helps me understand what needs attention during the shoot. I can also see what a motorized head would offer: consistent movements and less time between photos.

For the next shoot, it’s mainly the tripod I’d change.