Theory of Cell-by-Cell Color Perception in Oz Vision

Theory of color from cell to cell Barycentric coordinates are used to plot the colors produced by our Oz prototype on a Maxwell triangle (19). . This triangle projects its total activation (L, M, and S) while simultaneously displaying the color's chromaticity in two dimensions (2D). The natural human gamut, which encompasses all chromaticities that are attainable through normal spectral mixtures of light, is depicted in these diagrams as a color-filled subregion plotted at the bottom. In theory, the fundamentally larger color gamut that is accessible via cell-by-cell stimulation in Oz is the full area of the chromaticity triangle, assuming ideal conditions that produce perfect localization of light to target cones. However, in practice, a small amount of light will miss target cones and stimulate nearby cells, shifting the activation pattern away from the intended Oz color and toward the laser's natural color. In Fig., it is predicted how this stray cone activation will affect the colors that can be achieved. 2Opens in a picture viewer The point-spread function (PSF) of the laser microdoses on the retina in relation to the spacing of cone cells, the cone's spatial light-gathering function (20–22), errors in microdose targeting during eye movement, the proportion of L:M:S cones in the retina, and the stimulating wavelength are the most important factors. In the image viewer, Figure 2AOpens shows how the range of Oz chromaticities that are possible would be affected by fractional light leak. When stimulating the retina at 4° eccentricity, as shown, a diffraction-limited PSF would theoretically enable Oz to address nearly all possible chromaticities in the lms triangle, but not in the foveola, where cone cells are at their smallest. In reality, the total leakage of light includes more than just diffraction because of things like residual aberrations following adaptive optics focusing and microdose targeting errors caused by computational latency during eye movement. It is difficult to measure these factors directly, but the model shown in Fig. 2AOpens compared to the experimental color matching data in the upcoming "Color matching experiments" section suggest that two-thirds of the light captured by cones is captured by neighboring cones, while one-third is confined to the target cell. In spite of this unintentional light leak, our Oz prototype displays color beyond the normal human gamut with this level of accuracy. In the image viewer, Figure 2BOpens shows how the range of Oz chromaticities that could be achieved would be affected by the stimulating wavelength. Depending on the number of different cone types that respond to a given stimulating wavelength, the shape of this gamut can be a triangle, a line, or a single point. This shape reflects the relative responses of the L, M, and S cone cells. Prototype design An AOSLO (14), which simultaneously images and stimulates the retina using a raster scan of near-diffraction-limited laser light over a 0.9° square field of view, serves as the foundation for our Oz prototype. We are able to monitor the eye's movement in real time by imaging the retina with almost invisible infrared light. We dynamically target each cone cell in the field of view with pulses of visible-wavelength laser light to compensate for this motion. A population of 103 cones receives these laser microdoses at a rate of 105 per second. The spectral type of each cone needs to be known in order for cone-targeted stimulation to result in the intended LMS activation. Using recently developed optoretinography techniques in an AO-OCT system, cone cells in the subject's retina are categorized according to their spectral type in a preparatory step (15, 23). We use a classified region with 1000 to 2000 cones near a 4° eccentricity from the foveola for this study. We conduct the following experiments on human subjects: color matching with uniform Oz color squares and image/video recognition. The stimulated area falls within the classified region of the retina because all Oz stimuli are presented within the 0.9° square field of view, 4° adjacent to a gaze fixation point. Stimuli are randomly repeated in a control condition, and the microdose delivery is intentionally compromised. Each microdose in these control trials is randomly "jittered" so that it lands two cones away from the target. Studies on matching colors In order to officially measure the chromaticity coordinates of Oz colors, we carry out color matching experiments. Both 488 nm, which can activate all three L, M, and S cone types, and 543 nm, which is close to the peak of L and M but only minimally activates S, are tested as stimulation wavelengths. A red-green-blue (RGB) projector and a near-monochromatic laser with a tunable wavelength that can be mixed with white projector light are the two color matching systems we use. The latter is able to produce colors that are close to the natural human gamut, making it clear that our attempts to display olo do not actually go beyond the natural human gamut. The subject must judge match equality using the same patch of retina in order to eliminate effects from differences in adaptation across the retina because the subject sees 0.9° squares of Oz and controllable color that coincide in space and alternate in time during a color matching trial. Subjects can, as is typical with color matching (1), add light to the Oz color—also known as "negative" light—if necessary to achieve an exact match; the matched color is calculated by subtracting its color coordinates from those of the controllable square. In addition, subjects are asked to qualitatively name the hue and rate the saturation (on a scale of one to four) of the squares of controllable color and Oz color. The color matching experiments' graphs are shown in Figure 3 (opens in an image viewer). 222 color matches were performed by five subjects. We draw attention to four observations. First, around the stimulation wavelength of 488 nm, Oz colors form a triangle (Fig. 3BOpens in the image viewer), as well as a 543 nm color line (Fig. 3AOpens in the image viewer), which is in line with the theory in the section titled "Theory of cell-by-cell color." Second, as anticipated, the color "collapses" toward the stimulation wavelength under the jitter control condition. Thirdly, the further Oz colors are from the color matching system's gamut, the greater the variance in matching lms chromaticity. This pattern is in line with the geometric analysis in the Materials and Methods section titled "Plotting perceptual uncertainty in matching," which explains why perceptual uncertainty in chromaticity rises when light must be added to the test color to match. Fourth, Picture The image viewer's 3COpens provide unambiguous evidence that olo goes beyond the normal human spectrum. To match the (nearly) monochromatic colors shown, which are on the boundary of the natural human gamut, all subjects found it necessary to desaturate olo with projector white in these matches. Under the test subjects' viewing conditions, these matching monochromatic wavelengths between 501 and 512 nm are the most saturated teal hues for normal color vision. Although the Abney effect—a shift in hue with saturation (24)—opens the possibility that the hue of the wavelength at best match may not precisely represent the hue of the undiluted olo color, subjects' qualitative hue naming and saturation ratings support these quantitative findings. The color names "teal," "green," "blue-greenish," and "green, a little blue" were all suggested for olo. Fig.'s near-monochromatic colors have an average saturation rating of 2.9, while subjects consistently give olo a saturation rating of 4. 3Creates in the image viewer. Experiments in image and video recognition In order to investigate the capacity of human subjects to comprehend Oz-generated images, we create image and video recognition experiments. We use four-alternative forced choice (4-AFC) and two-alternative forced choice (AFC) tasks in which participants can only succeed if they use accurate Oz stimulation-generated hue information. Subjects must determine the image's line orientation in the 4-AFC task. Subjects must identify the direction of rotation in a video of a moving disk in the 2-AFC task. With a stimulating wavelength of 543 nm, these stimuli present the lines and disks as red (all-L cone) on an olo background (all-M cone). In the jitter control condition, all hue and luminance cues are eliminated and the task is reduced to guessing after a calibration step is carried out (see the "Image and video recognition experiments" section in Materials and Methods). The results of the 4-AFC line orientation and 2-AFC rotation direction tasks are plotted in Figure 4Opens in an image viewer. Subjects are able to accurately detect both line orientation and motion direction (blue bars) in the experimental condition. Subjects' performance in the jitter control condition is reduced to guessing for both tasks (gray bars). When the task was easy, subjects report seeing red or orange lines and disks on a blue-green or green background, whereas when they were forced to guess, they see a yellow-green square. The first is directly related to precise Oz microdose delivery, and the second is related to the jitter control condition, in which only the natural color of the 543-nm light should be seen. DISCUSSION

Today's color reproduction technology, such as CMYK printers and RGB displays, is based on spectral metamerism, which produces light with the same activation level as a target color for each type of cone in the retina. This method has been used since at least 1861, when Maxwell demonstrated live at the Royal Institution how to create full-color images by superimposing red, green, and blue images (25). In this paper, the Oz principle of color reproduction is

WEITTED BY ME