Before starting work on anything, I had to think about what it was that I was after exactly. After some head-scratching, and lots of doodling, I came up with the following problem definition:

Develop a method to split up an image into multiple, varying-sized, non-overlapping squares in such a way that the sizes of the squares can reflect an underlying image in some meaningful way.

This became the guideline for a simple program I developed over months to do just that. The image to the left is an early result. Though it is a little rough, this image was enough to let me know that I was on the right track. I could now proceed to the next stage, namely populating those empty squares with actual images.

Creating a visually pleasing mosaic proved to be almost as challenging as coming up with the tiling algorithm itself. The image on the left is the result of one of my earliest attempts.

To generate this image, I took a couple of hundred images from my film school thesis, and randomly populated the tiled spiral data set shown above. The result is, needless to say, not easy on the eyes. Nevertheless, I this image is special to me because it is at once a snapshot of my earlier film work, and a promising start to a new art form.

What if, instead of tiling up a bunch of images to generate a composite like the one above, I use the same tile data to alter a single image in some way? An answer appears to the left.

The mosaic I created from this image is the result of introducing random variations of hue and brightness within each individual tile using the same spiral dataset as in the previous two images.

I took the original photograph for this mosaic during a trip to Paris a few years back. Its subject is a statue of Louis XIV found on the grounds of the Palace of Versailles.

After weeks of experimenting with different tiling methods, I arrived at the image on the left. The underlying photograph is of a beloved Los Angeles landmark currently undergoing a much needed renovation: The Griffith Observatory.

Although the observatory itself is barely discernible, this image helped me understand that the more complex a mosaic composition, the simpler the individual elements that make up that composition need to be. Otherwise, you end up with a lot of noise that's just hard to look at.

In this mosaic, the individual elements that make up the observatory and the LA skyline are just stars--thousands of them.

This mosaic has three distinct new features:

1. Tiles overlap, thus creating a sort of scaly effect.
2. Tiles are composed of two different sets of images: clouds for the sky, and a cracked earth texture for the ground.
3. Tiles vary in size according to distance from the viewer

In the not-too distant future, we will have cameras that will give us "distance" in such a way that we will be able to take three-dimensional pictures. I think that this is more or less what I was trying to do with this piece, namely enhancing the concept of distance within the image by sizing tiles accordingly.

This composition was inspired by traditional Japanese woodblock prints, and the work of the Russian-born painter Erte.

Once again, this mosaic started off as a simple pencil drawing. With some modifications to the tile--generating algorithm, I was able to generate tiles that flow along the contours of the waves.