Unlike text data, video in computers has no intrinsic semantics. Rather, it is saved as flux of rectangular pixels and there is no standard known or used worldwide to represent it’s contents that is analogous to languages like, say, HTML for text. Therefore, a computer that has no aditional programming is not capable of answering such simple questions as wether a video contains a person or an object, if it is moving or if it is lighted. Computer vision was developed to solve these problems.
Three elementary techinques help distinguish which pixels in a video signal belong to people or objects: frame differencing, which finds the object of interest by detecting it’s movements; background substraction, which identifies visiting pixels by comparing them with a known background and brightness threshold, which uses expected diferences in brightness between the foreground and the obect of interest regarding it´s background. These methods are easy to implement and help contitute the basis of detection schemes with which more complex systems of interaction can be created.
Mirrorwall was constructed as an exploration of these forms of interactivity, and I chose frame differencing because it seemed the most useful techinque for an environment in which I would have the opportunity to showcase it: the presentation of Abie Toiber´s album Relativity, a crowded event where everyone was moving. We had limited access to lighting, but fortunately the detection worked with even better results than we had foreseen.
At first it was projected on a white wall in a rest area in The Lunario of Auditorio Nacional, and it generated images based on the movement of the peolple who arrived to the concert. Later, during the concert, it was shown on screens that acompannied the band’s scenery, and generated images based on the movements of the band while playing.
My intention was to create an analog impresión to the sucesion of real events and their signification through a subjective process, under the idea of a sort of illusion in which objects and forms exist because we percieve their permanence throughout the passing of time.
Pictures from the application itself.
(Click the thumbnails for more detail)
 "giro_01.gif")
 "giro_02.gif")
 "giro_03.gif")
 "fuego_01.gif")
 "fuego_02.gif")
 "fuego_03.gif")
 "carita_feliz.gif")
 "portada.gif")
Mirrorwall was programmed using Processing (version 0124) with the help of the video library for the implementing movement detection algorithms. The traer.physics library was used for particle generation.
The algorithms used for image processing and particle generation were coded with the help of Víctor Valenzuela - murmullo.net.
For testing and development we used two computers and three cameras:
For the final presentation we used the miniDV camera because there was some lag using the others. Projection took place thanks to Don Chakra Corona's generosity, who lent us his projector.
with the imagery created by Mirrorwall, while I was carrying the camera.
This version of Mirrorwall presented at El Lunario contains a small modification from the original version wich includes fragments from Abie Toiber's songs.
Pictures were taken by Víctor y Gerardo Guerra Pilón.
Thanks to everyone!
Pictures of Mirrorwall in the presentation
(Click the thumbnails for more detail)
 "portada_envivo.jpg")
 "camara.jpg")
 "silueta.jpg")
 "perestroika.jpg")
 "chamuco.jpg")
 "baile.jpg")
 "metro.jpg")
 "dedos.jpg")
 "mano.jpg")
 "desfase.jpg")
 "grenhas.jpg")
 "abie.jpg")
 "corte.jpg")
 "semanatres.jpg")
 "mezcla.jpg")
 "humo.jpg")
 "rastro.jpg")
 "concurrido.jpg")
 "vereda.jpg")
 "paola.jpg")
 "de_espaldas.jpg")
 "1981.gif")