Human vision is a complex thing. Light enters through a pair of not very high-resolution ‘cameras’ positioned a few centimeters apart and seldom identical. The light is transmitted to the visual cortex at the back of the brain by nerves activated by electrical impulse. This action potential seems to be triggered by changes in local light levels with the nerve cells acting like individual pixels in a composite image. The eyes are constantly in motion informing the response in our brain as it creates the illusion of an image from a mosaic of glances, memorized information, peripheral vision, orientation, registration, repetition – vision is dependent on movement and the visual cortex is making sense of a vast amount of data as it forms a multilayered ‘picture’ of the things we see. Human vision is dependent on a mix of optics, memory, compilation and processing. It is noisy, located, perspectival and subject to error. Perhaps the computer vision strategy used by Franquelo is more ‘true’. It is still located but the computer controlled ‘pan and tilt’ head carrying the optical system revolves around a single known point. A focal ratio is a dimensionless number but a low f-number produces a sharp image and a shallow depth of field. By repeatedly re-focusing and rephotographing multiple captures of the same target will contain different information - from this pool of data it is then a question of selection. The selection involves the blending of the pixels from different focal planes by maximizing the contrast with the surrounding pixels. Once identified, these ‘sharp’ pixels are merged to form a composite from which both 3D data and image data can be extracted. The algorithms which control these actions have their own poetry. The more elegant the algorithm the closer the relationship between the object and the image.
Extracted by Adam Lowe from
conversations with Manuel Franquelo that started in 1996