Rethinking the Process of Vision

Vision science must stop thinking of the retina of the eye as the image plane of a camera or as the analogue of photographic film or the array structures used by contemporary silicon (CCD or CMOS) imaging chips!

A condensing lens (as  in the eye )  has two  image planes.  The nearest is the focal or Fourier plane where image information is encoded in two terms as light intensity and phase, and, the “image” plane that is sensitive only to light intensity. The latter is the point where film is placed in a camera and forms the basis for most of contemporary imaging technology. The eye, however, evolved to image in the Fourier domain that never seems to have been recognized in a hundred years of vision science

See the retina as the traditional, intensity-only sensitive, image plane  is  fundamentally at variance with the historical measurements  of retinal morphology made by Osterberg in  1935  (recognized and quoted ad infinitum  in the literature of vision) of the spatial arrangement of the rod and cone receptors of the retina. It is just  illogical to persist in this thinking.

It is seen from this work that  from a simple geometrical premise the retina is a diffractive array  of receptors that is demonstrably the focal or Fourier plane of the optics of the eye. Further, I have proposed retinal devices that are capable of encoding  both light intensity and phase as required by the Fourier equation.

The  diffractive pattern of light detection on the retina is  not the uniform array of “red/green/blue detection centers” that would be required by film  Moreover, it is  clear that this pattern  derives (or evolved) directly from the chromatic aberration of the lens of the eye where different wavelengths of light are refracted through different paths within the lens to different positions on the focal or Fourier plane behind the lens. THIS WORK DEFINES WITH GEOMETRICAL PRECISION WHERE THE WAVELENGTHS OF LIGHT WITHIN THE VISUAL BAND FALL ON THE RETINAL SURFACE FROM THE FOVEA TO THE PERIPHERAL RETINA.

The vision process is in the “Fourier” domain and not the “intensity only” domain of the film camera. This means that the individual light detection centers of the retina possess the capability for encoding both light intensity and phase as required by the Fourier equation. Phase encoding implies that the direction of the impinging light ray is retained and can be used eventually to process the image. In a sense the Fourier plane is the most efficient location to encode the image – a property that nature would certainly make use of.

Moreover, the plan of light interaction on the retina undoubtedly directly reflects the, heretofore termed chromatic “aberration”, of the solar spectrum transmitted through a condensing lens. It becomes clear (at least to me) that the evolution of the eye derives beginning somewhere way back there when a combination of sunlight incident on that first condensing lens (a drop of water?) and the molecular self-organizing character of some polar long chain hydrocarbon molecules that happened to be in the vicinity. All that was needed beyond that were the physical laws of light refraction…and…in due course of a long long time…….the eye and vision!

This explanation opens a great many new vistas in vision science. Perhaps a critical insight is that an exact mid-point of the visual band is defined on the retinal surface is geometrically-defined! This would correspond to 550 nm if the band is assumed to extend from 400 to 700 nm. No “laboratory spectrometer” is needed. This wavelength – defined as the “green” primary color – is determined by the geometric ratio of the size of the cones and rods. It provides the genetically determined (through the sizes of cones and rods) “universal reference standard” where all of us “see the same green”. This fixed reference is also used in retinal computation to determine the many hues of color that we see. The existence of such a fixed reference was presciently and brilliantly seen by Edwin Land from measurements made external to the eye. He termed it the “fulcrum” of retinal response.

Beyond this, the work provides an explanation for the remarkable (and never explained) ability of the eye to sense single quanta (or photons). I have written about this – see previous Comments. In summary, the vision process, at the point of light interaction with the retina, will be seen to lie in the spatial and temporal domains of quantum physics. Details cannot be presently seen but the beginnings of an understanding may lie in the recent Comment proposing the formation of standing waves within retinal receptors (or, alternatively, within the inter receptor devices that I propose) involving the frequency of light. I would note, again as I have written, that any fast, quantum processes on the retina are probably slowed to “human nerve scale” by the lengthy transmission of image information through the lengthy optic nerve.

I have been reading “The Silicon Eye” by George Gilder (Norton, 2005) recounting developmental work on silicon imaging chips that are based to some extent at least on replicating the light response of the human retina. I will have more to say about this shortly. It seems, however, that the fundamental basis of the effort is the assumption, again, that the retina is the analogue of photographic film, i.e., intensity-only detection! Beyond this, the work aims at besting contemporary silicon imaging chips by designing each single light receptive pixel space to process the three primary RGB colors instead of the separate R, G, and B, spaces employed by contemporary imaging (CCD or CMOS) chips (“that waste two thirds of the incident light”). The impression is given that these new “single RGB sensitive pixel” imaging methods approach the behavior of the human eye. I would propose that this is not nearly the case.

The “Fourier retina” of this work is divided into precisely three RGB wavelength-tuned regions each of which is composed of an array of single devices that are capable of detecting, and processing a usable electronic signal from, single quanta interactions. I have proposed the structure of, and the physics underlying, these devices. This, in turn, implies that the quantum efficiency within each region is very high – near unity. Moreover, the density of individual receptors in these arrays approximates one hundred million per square centimeter, i.e., one hundred “megapixels”. I think that we have much further to go before we replicate the biological retina on silicon – but I know the first steps to take!

GCH

Tucson, AZ

11/4/07

This entry was posted on Sunday, November 4th, 2007 at 2:27 pm and is filed under Running Commentary . You can follow any responses to this entry through the RSS 2.0 feed. You can leave a comment, or trackback from your own site.