Human visual perception

The most important function of visual processing is to reconstruct equivalent representations of the world. The ability of the visual system to represent surfaces is fundamental to visual perception. Human visual system processes informations in three levels: on low-level processing photoreceptors are detecting the stimulus, the horizontal, bipolar and amacrin cells are pre-processing the sign, ganglion cells are submitting them via optic nerve into the brain.

On mid-level processing contours are grouped into images. On high-level processing brain interprets viewed objects. Brightness perception is concerned all in the three levels [6].

The cone photoreceptors represent the initial fundamental sampling step in the acquisition of visual information. The human photoreceptors are the rods and cones in our retina, which send the stimulus by neural networks to the visual cortex, where brightness perceptions are formed in. Studying the structural properties of the normal cone photoreceptor mosaic is important to evaluate how the human visual system samples the world. The receptors of colours are the cones situated in central part of retina, in the fovea. There are three types of cones sensitive to three different spectra: L for long, M for medium and S for short wavelengths. Conventionally S cone is blue-sensitive, M cone is green-sensitive and L cone is red-sensitive. The S cones are unique among the cones: they are only about 2% of the total number and are found outside the fovea centralis where the green and red cones are concentrated. The center of the fovea is the foveola – about 0.2 mm in diameter – where only cone photoreceptors are present and there are no rods. Ganglion cells collect the informations from rods and cones. Ganglion cells have circular receptive fields, with centre-surround opposition, most ganglion cells have either ON-centre OFF-surround receptive fields or the reverse. The small parvocellular ganglion cells, or P cells, represents about 90% of the total population of ganglion cells, and handle the majority of information about colour. The response to a given wavelength at the centre of their receptive fields is inhibited by the response to another wavelength in the surround. P cells fall into two groups: one that processes information about differences between firing of L and M cones, the red-green channel; and one that processes differences between S cones and a combined signal from both L and M cones, the blue-yellow channel (opponent colour theory). The ganglion cells' axons form the optic nerve and thereby transmit information from the retina to the lateral geniculate nucleous (LGN) and after send the stimulus by neural networks to the primary visual cortex (V1) located at the back of the brain. Colour perceptions are formed in our brain [7].

Wilks et al [8] said that the location of the different foveal specializations is variably offset from the PRL (preferred retinal locus) and according to their results pit volume is correlated with FAZ (foveal avascular zone) area, but not with peak cone density.

Elsner et al [9]'s results indicate that the variability of total cones across a large portion of the macula is much less among individuals, when the eye is healthy, eye length is normal, and age

is restricted to young adult ages. Many individuals with the highest cone densities near the fovea have the lowest cone densities in the periphery.

According to Rees et al [10], parafoveal function is important for daily visual tasks such as reading. The variability in cone density along the four cardinal meridians in parafoveal regions of the retina was investigated in vivo using an adaptive optics fundus camera. Ten healthy normal trichromatic individuals were included in the study. There were significant differences in cone density between individuals at all four tested eccentricities (0.5, 1, 2 and 3°) and meridians.

Roe et al [11] investigated monkeys and they found that luminance responsive cells are located in color-activated regions of primary visual cortex (V1), whereas Cornsweet responsive cells are found preferentially in the color-activated regions (thin stripes) of second visual area (V2). This colocalization of brightness and color processing within V1 and V2 suggests a segregation of contour and surface processing in early visual pathways and a hierarchy of brightness information processing from V1 to V2 in monkeys.

The perception of brightness can not be measured directly, human visual system has some uncertainty to judge the brightness. Luminance can evoke different brightness perceptions and several attributes (temporal, spatial, etc) can influence the brightness. The International Comission on Illumination (CIE) prepared a photometry system in 1924 for measuring brightness by a visibility function V(λ) [12].

Human vision is not merely a simple light measuring device: an object can be seen in relation to its environment, thus simultaneous effects are important. We need discover more knowledge about human vision for promoting optimal perception.

Large homogeneous coloured objects are perceived homogeneous both in colour and brightness, despite of changing the retinal colour sensitivity with the excentricity. Appearance of large samples are dominated by local peripherial colours, and these are independent from contours or from blurred boundaries in the viewed image [13].

The colour of a surface changes when it is surrounded by other colours. Webster presents a new illusion to show that these changes are much stronger when seen through other colours as transparency, suggesting that the brain parses the causes of colour into separate layers [14].

Gilchrist [15] showed that the lightness of a surface depends mostly on where it seems to lie amongst other surfaces. A dimly lit piece of paper looked white when perceived to lie in the same depth plane as a dimly lit black piece of paper, but it looked almost black when it seemed to lie in the same depth plane as a brightly lit white paper. In Gilchrist's ingenious demonstration, the retinal image formed by the paper and its surround was exactly the same in the two instances (only the viewing conditions were manipulated). His results thus imply that, for lightness perception, the visual system computes the contrast only between retinal areas that belong to the same depth plane.

Wallach [16] proposed a simple ratio theory of lightness. Presenting observers with two disk-annulus displays, he showed that disks of different luminance appear equal in lightness as long as the disk-annulus luminance ratios are equal.

Helmholtz observed that the effects of contrast could be dramatically reduced by enclosing the grey patch in a black contour that "sharply divide(s) it from the ground". His results showed that the dividing contour might not eliminate totally the contrast effects, but argued that the patch could now be 'recognized' as pure grey [17].

White's illusion [18] presents two objects of equal luminance which seem differently: the gray bars that appear lighter are those that are surrounded mostly by white while the ones that appear darker are mostly surrounded by black. (Figure 1.)

Figure 1. White illusion

The watercolor illusion, or water-color effect, it can be seen on Figure 2. shows that the vertical gratings are black and white with a thin line of red along each black bar. The horizontal gratings are black and white with a thin line of green along each black bar. The illusion is that the red and green appear to spread over the black and white regions of the vertical and horizontal gratings respectively. Pinna, [19]

Figure 2. Watercolour illusion

Several brightness illusions indicate that borders can affect the perception of surfaces dramatically. In the Cornsweet illusion, two equiluminant surfaces appear to be different in brightness because of the contrast border between them. In Figure 3., left part of the picture seems to be darker than the right one. In fact they have the same brightness. [20]

Figure 3. Cornsweet illusion

The perception of surface brightness is influenced not only by local surface luminance but also by luminance and border contrast cues in the surrounding scene. After Weil [21] the study of fading and filling-in is relevant for a better understanding of brightness, color, and texture perception on large uniform areas. Without filling-in we would see just borders, no surfaces.

Perceptual filling-in occurs when structures of the visual system interpolate information across regions of visual space where that information is physically absent. Perceptual filling-in is the filling-interpolation of missfilling-ing filling-information across visual space. It is a ubiquitous process filling-in the central visual system, necessary to make sense of the world. Light from portions of objects and scenes often falls upon parts of the retina without photoreceptors, such as the

blind spot or retinal vessels, or falls behind objects in the real world and the visual system must process information across these occluders so that they are perceived as complete and not fragments. It is an extremely effective process, as most of the time, we are entirely unaware that it is taking place. It is also a natural process that takes place all the time, but researchers have observed that certain stimuli can be configured to promote perceptual filling-in.

Results of Li et al [22] suggest that the strength of filling-in decreases with distance from the fovea consistent with the decrease of the cortical magnification factor.

After Anstis [23], if a red disk is surrounded by a green annulus, and the border of the red disk, but not of the green annulus, is retinally stabilized, then the red disk is gradually filled-in perceptually with the green color of the annulus, and the whole field looks green. Although the red disk is still present in the field, it becomes invisible. There are two theories of how this might be done: isomorphic or neural filling-in, and symbolic filling-in.

1. According to the isomorphic theory, a surface to be filled-in is represented in the brain by a two-dimensional array of neurons, and these actually fire under the influence of excitation spreading in from the edges. It is suggested that color signals spread in all directions until they are stopped by a luminance border that acts as a barrier. This process is thought to be analogous to physical diffusion.

2. By the alternative theory, symbolic filling-in, there is no spread of activity within the surface. Instead, visual properties of the surround, such as texture, contrast polarity and color, are tagged and applied to the enclosed surface. This process is sparse and economical and resembles the vector graphics method of representation. It is not clear how this would be implemented in the nervous system.

As we can read from Von der Heydt et al [24], information about the color and brightness of a surface is represented in two ways in the visual cortex, by the activity of neurons whose receptive fields point at the surface, and by the responses of neurons whose receptive fields straddle the border of the surface. Recordings from the visual cortex of monkeys during perceptual filling-in failed to show the corresponding change of the surface activity. They conclude that the physiological evidence is incompatible with an isomorphic filling-in theory which assumes that color signals spread from the borders into uniform regions. We have used pairs of complementary colors for disk and ring (and switched them between trials to keep color adaptation constant). However, informal observations showed that similar filling-in would be obtained with other colors, for example, a gray patch surrounded by a colored ring or vice versa.

Sasaki says that [25] it is presumed that surface representation is produced mainly in the midlevel vision and that area V1 (the primary visual cortex) activity is solely due to feedback from the midlevel stage. Activity for filling-in was observed only in V1, whereas activity for illusory contours was observed in multiple visual areas. These results indicate that surface representation is produced by multiple rather than single processing.

In the experiment of Hamburger et al [26] a central disk and two concentric rings were used as well as similar stimuli consisting of three nested squares or parallel stripes. They tested filling-in with different equiluminant colour combinations, and observed four modes of filling-in: First, in most of the cases, the inner ring assumed the colour of the central disk and outer ring (M1). Second, the central disk became filled-in with the colour of the inner ring, without any colour change on the outer ring (M2). Third, in a first step, the colour of the inner ring spread onto the central disk; then, in a second step, the colour of the outer ring spread over the whole stimulus (M3). This two step filling-in process has not been reported so far.

Fourth, a mode (M4) was sometimes observed that was characterised by the central disk and outer ring assuming the colour of the inner ring. Thus, colour filling-in or colour spreading proceeded both in a centripetal (periphery to fovea) as well as a centrifugal direction. The

colours red and yellow proved to be stronger inducers than blue and green. Conversely, the latter colours became filled-in more easily than the former. The filled-in colour was always that of the inducing stimulus, i.e., there was no colour mixture. This suggests a long-range, neural process underlying filling-in under these conditions.

Subjects predominantly reported perceiving the inner ring as figure and the central disk and outer ring as ground. Rather than seeing three individual stimulus components, they perceived a coloured figure on a grey background. Under these conditions the ring faded into the background. Red and yellow showed themselves as strong inducers, whereas blue and green were more susceptible to becoming filled-in. A possible interpretation for the different strength of colours is that blue and green usually resemble background colours in natural scenes, whereas red and yellow are typically associated with properties of objects (foreground), such as the ripeness of fruits. Therefore, they would be expected to have a higher perceptual salience.

The grey inner ring became much more frequently filled-in by the colour of the perceived background (central disk and outer ring) than the coloured inner ring before. It thus appears that grey is not only a weaker inducer than each of the four colours tested, but is also more susceptible to filling-in. This assumption is consistent with the results showing that all subjects perceived the grey inner ring as figure and the equally coloured disk and outer ring as ground.

Sasaki [27] found that the activity in the color filled-in region was observed only in primary visual cortex area V1 when attention was controlled, whereas the activity in both illusory and real contours was not confined to V1. These results suggest that surface representation is not a result of single processing but of more complex multiple processing. The present results show that contours and the filling-in of a surface feature are processed separately.

Hamburger et al [28] specifies the conditions under which fading and filling-in can occur. It is well known that with prolonged fixation targets assimilate into the background and become invisible, Such backgrounds need not be uniform and steady; a textured background or dynamic visual noise are as effective and can be even more so. Two hypotheses may account for fading and filling-in:

(i) Generalization of brightness or texture across the enclosed surface region by edge-selective cells using form and color information; and

(ii) active spreading of information from the edge by way of lateral propagation.

There is evidence for both in cortical areas V1-V3.They have found that filling-in requires little surround information. For example, a thin red ring hugging the boundary of the physiological blind spot will uniformly and completely fill-in the enclosed blind spot area.

Similarly, a thin chromatic double contour will induce watercolor spreading over a wide area.

These observations suggest cortical mechanisms involving long-range horizontal interactions to account for brightness and color perception on uniform areas.

Microsaccades are important for visual perception, (i) because they refresh the border signal and thereby prevent stimulus contours from fading. Furthermore, (ii) they also help to sustain the brightness of the interior by spreading contour and edge information laterally. Perceptual filling-in enables us to perceive uniform brightness, color, and texture on extended surfaces that would otherwise quickly deteriorate, leaving us with contours, but no surfaces. The same long-range mechanism that fills in and sustains a target from the boundary presumably also fills in the target area with features of the background, when the contour breaks down because of the lack of eye movements.

After Troxler [29] effect, in case the eyes fixate a point very steadily, small objects in peripheral vision often fade out from view and disappear. [29]

According to Grossberg et al [30], figure-ground perception enables us to perceive objects that are distinct from one another and from their scenic background. Many factors contribute

to figure-ground separation, including differences in luminance, color, size, binocular disparity, and motion between a figure and its background. Image size alone is not a reliable cue to a figure's depth. In particular, a nearby small object and a far away large object may both subtend the same \size" on the retina.