Ken Salter

Reflecting Pool


Participants interact with the Reflecting Pool using gestures.  Slow and deliberate gestures, in the style of Tai Chi hand movements, work best. Also periodically dropping one's hands to the hips allows the images to fully develop their form, much like ripples settling on the surface of a pond.  Basic gestures include:

  1. Tilting hands upward or inward - causes mosaic windows to zoom in (larger)
  2. Moving hands sideways in the same direction- causes the entire mosaic to rotate.
  3. Moving hands sideways in opposite directions - causes a kaleidoscopic rotation.
  4. Moving one hand up and the other down - causes twirls.

Gestures depend on time.  The longer a gesture is maintained, the greater an effect.

Mosaic of Monas (1st pass thur video feedback loop)
Mosaic of Monas (1st pass thur video feedback loop)
Mosaic of a Mosaic of Monas (2nd pass thru video feedback loop)
Mosaic of a Mosaic of Monas (2nd pass thru video feedback loop)


The Reflecting Pool is part photography, part algorithm/digital processing.  The device consists of three main parts which, together, form a video feedback loop:

  1. A horizontal, upward facing video screen.
  2. A downward facing camera pointed at the video screen.
  3. A computer which constructs a mosaic of images out of the camera image and sends the mosaic to the video screen.

The result is that for each video frame, the camera takes a picture of the previous mosaic.  The next frame is a mosaic of mosaics.  This repeats, over and over again, causing an infinite nested pattern.

The simplicity of the algorithm can be seen in the mosaic shown on the first image with twelve Mona Lisas.  A second pass begins to establish a complex pattern, as can be seen in the second image with 144 Mona Lisas.

The mosaic consists of transformations (rotation, zoom and translation).  The gestures affect the values for the transformations.


Emergence of Color from Black and White


In the adjacent clip, Reflecting Pool is seeded with an initial image: a black and white checkerboard.    For the first several iterations, the results are what you might expect: a black and white image with ever increasing fine detail.

But after the first several iterations something remarkable happens.   Colors spontaneously emerge and the detail coalesces into coherent forms and patterns.  Also surprisingly, these forms gracefully morph into different forms despite the fact that the algorithm, and all of its parameters, are unchanging.  This is the “strange” part of the loop:  the lower levels of detail somehow coordinate with one another to make a coherent big picture.