A large aquarium shot from below the water line. The trees above the water change the light over the course of the shot, and some large animals are disturbing the water, also changing the character of the water. Here the compressor is separating clear lines as the light in the water changes over time, suggesting that the water image is not continuous. This is particularly strange in water imagery as we don't expect it to be segmented in this way.
The distortions in this image are subtle, and it appears to be a normal panorama until you look carefully at the short grasses on the left side. There you will see details of the image attempting to knit the sere blades into one. The attempt is brave, and futile, though the result is generally quite fluid.
This beautiful texture emerged when I asked a digital camera to sort out imagery structured by a regular gridded screen installed behind colored, textured glass.
Time multiplied these trombones into space.
The sunset stole from the compressor's idea of the subject of this image and made short work of many people in the frame. See how they have been cut and multiplied, their images assembled from many points in time as if they were separate people. Yet the background retains much of its continuity and the colors remain rich and appealing.
A compressor challenge much like *Kitchen Floor,* though stranger. I recorded my feet walking on a bridge across the Los Angeles river. The diamond plate forms an irresistible grid, less liney than kitchen tile but equally attractive to the compressor. Shadows of
This early sketch shows the compressor trying to place the subject (my feet) in context of a gridded vinyl floor. It's a clear failure, and it tells a story of how MPEG-4 may segment background **and** subject incorrectly.
research. tests. these are examples of MPEG-4 grouping imagery together, reducing motion to stills. ultimately these will be explored as animations.
these video panoramas show MPEG-4 working out the boundaries of their subjects. the results are consistent enough to start using this process. the variables are speed of motion, brightness of subjects, "griddiness" of the subject, and distance from the sensor.