Early Visual Processing PDF

Early visual processing Sensation: refers to physical responses to the environment by our sensory organs and subsequent neural systems → sensory information is detected by a sensory receptor, environment signal turned into neural response Perception: efers to the way sensory information is organize...

Early visual processing Sensation: refers to physical responses to the environment by our sensory organs and subsequent neural systems → sensory information is detected by a sensory receptor, environment signal turned into neural response Perception: efers to the way sensory information is organized, interpreted, and consciously experienced, the way we interpret and experience sensation. Difficult to measure directly, but also involves the brain. Simplified, created by the brain. Color constancy, shortcuts. Memory fills in visual acuity gaps. Bottom-up processing: perceptions are built from sensory input. Top-down processing: how we interpret those sensations is influenced by our available knowledge, our experiences, and our thoughts. Perception =/= sensation =/= perfect representation of the outside world On the eye level: Light from a point scatters, and needs to be brought back to a point on the retina (cornea +lens). The cornea affects vision in multiple ways: It bends most of the light (more than the lens) and it filters out UV (short wavelength)-light. The second thing we run into that has direct affects on vision is the iris. It affects the size of the pupil. Pupil varies the amount of light the enters the eye. A lot of lights seems great but isn’t always. The photoreceptors are very sensitive but also, a large pupil has a large edge and the edge messes with the bending of the light (=lower acuity vision). Lens and cornea: bend light, you need to invert the scatter, to see things with acuity. Curving the surface: changes the relative response between A and B, angulation of the light source gives different responses in cells. Helps you figure out the source of the light in space. It is all about relative responses: never works in isolation. Retina: optic nerve (signal travels), Photoreceptors interact with the world. Spatial location is filled in by the other eye, blindspot. Brain reconstructs. Human vs Octopus: photoreceptors on the front, inverted. Blood supply is not as good them. Here, humans are the strange ones. In octopus, the photoreceptor outer segment, which interacts with the photons, faces the lens. In humans, it faces away into the retina, and nerve fibers and processing go back towards the middle of the eye. So out retinal processing blocks some of the light coming through. This creates a blind spot where the nerve fibers pass back through the photoreceptor layer and to the brain. This is a bad design, but that’s just how it evolved.  Eccentricity: Distance from the fovea (highest density of photoreceptors) for location on the retina. Higher eccentricity→ more towards the periphery of your vision. 0=fovea. Topographic map properties: Center of visual field represented by more neurons than periphery. Neural representation is again distorted, to over-represent the central visual field. But neighbourhood relationships are still maintained. The spatial arrangement has direct effects on perception (acuity & color vision but also motion direction). Stimuli for vision: 400-750 nanometers for humans. Different perception. Main difference between rods and cones is to what wavelength they are receptive. Cones: 4 types in women, 3 in men. Concentrated on the fovea. Rods: night vision, bleached during the day (pigments are broken down). Photoreceptors form the interface between the physical stimulus and the nervous system: These fine lines are layers of pigment that interact with incoming photons. Here, light affects the cell and modulates neural activity (transduction). The rod and cones are depolarised when not stimulated by light. This is due to high concentrations of second messenger, like molecule referred to as Cyclic Guanosine MonoPhosphate (cGMP; opens sodium channels). Normally the receptors have high cGMP levels that keep the sodium channels open, depolarising the cell. Cascade: The pigment in rods, rhodopsin, has 2 parts: retinal and opsin. The same holds for the pigments in the cone types but the shape of the pigment molecules is different. The main difference is the different types of opsin for different types of photoreceptors. 1. Normally, rhodopsin is bound to the discs in the rod. Light absorption with correct wavelength breaks the bond between retinal and opsin, resulting in the pigment basically falling apart (bleached, it becomes ineffective).  2. As such, the pigment is unable to fit to the binding site on the disc. The retinal part is now transported to other parts in the cell for recycling. This is important because we can't create retinal ourselves. But you can eat it (vitamin A). 3. Opsin starts second messenger that will decrease cGMP When light hits the receptors, the rhodopsin is broken down and the cell hyperpolarizes: 1. The activation, and thus destruction, of the pigment also sets in motion a cascade of events that leads to the break down of cGMP. It causes neurotransmitter release, Intensity coded in amount of neurotransmitter released. 2. In darkness cGMP concentration is high which keeps sodium channels open causing sodium to freely flow into the cell: leaving it depolarized. 3. When cGMP concentration goes down during light, these channels close. The result is hyper-polarization, with less response to light. 4. The principle is the same in our four cone types but the pigment gets broken up by other wavelengths → System responds to darkness and not light: neurotransmitter release in darkness. Photoreceptor Tuning Functions: full-width half max. If narrow, good at discriminating things, only responds to few stimuli. Blue cones (relatively low wavelength preference) then Rods, then green cones and then red cones. Cones work in a similar way, rhodopsin has a different shape and thus responds differently. Rods break down way more, they become completely bleached because they are overexposed but the cones can become renewed continually. They have different foldings/shapes. Pigments are constantly recycling in cones (homeostasis), but rods stay bleached the entire day. Red light for night vision: Rhodopsin in the human rods is insensitive to the longer red wavelengths, so traditionally many people use red light to help preserve night vision. Red light only slowly depletes the rhodopsin stores in the rods, and instead is viewed by the red-sensitive cone cells. Using a low intensity red light or green light helps preserve your night vision. It shortens the recovery time once you turn off white light illumination and leaves the eye’s night vision ready once the low intensity light is turned off. Night vision is in green: It is generally considered that red breaks down rhodopsin more slowly and, if preserving night vision is the main objective, red is better. But green light penetrates a little better, and shows more detail. It may be preferred for distance vision, and for close up clarity, such as reading instruments or maps. Green is more commonly used in military situations, where it is claimed to be less detectable by night vision equipment. Cones lead to color discrimination: more cones, more discrimination (and thus more colors) Female extra L-cone: Red cone is on X chromosome, can discriminate shades of red that are not visible to men. Causes relative response differences, ability to discriminated. Humans vs Mantis Shrimp: they discriminate better, they have more cones. More differences in wavelength. Receptive fields: location in any type of cell to which a cell responds to. If light comes from a location, the cone changes its response. There is an anatomical chemical receptor field in the brain but responds to the outer world. For a photoreceptor, the region of space in the visual field where light can come from and still excite the cell is referred to as its receptive field (RF). Note that RFs for vision are usually talked about in terms of the position in the visual field. But photoreceptor are of course only the first step. Before the signals leave the eye via the retinal ganglion cells, the outputs will first affect each other and also form new, more complex receptive fields Convergence: From -130 million photoreceptors to -1.25 million retinal ganglion cells, alot of input to less output. Purpose: Information reduction, increasing information complexity. Detect changes, disregarding constants. 1. Horizontal cells: First step after photoreceptors are horizontal cells. Connect to other photoreceptors. Enormous reduction in information through the retina: far less cells carrying more complex but compressed information. Reduction of information occurs through detecting signalling changes in the image and disregarding constant parts. Same method of compression is used in digital data: DVDs, MP3, JPG etc all achieve compression by signalling changes over space and time Lateral interactions: Activation (or lack of activation) from horizontal cells ‘spills over’ to nearby photoreceptors (making them either more or less active) in cell activity modulation → activity in horizontal cells generalizes in a larger effect than expected. → In dark: receptors depolarized, photoreceptors depolarize horizontal cells, which release GABA (inhibitory). Horizontal cells hyperpolarize photoreceptors. When light hits photoreceptor, it depolarizes further. On Bipolar cell will depolarize, release a lot of neurotransmitter. Center-surround on-centre off-surround will be stimulated. Mach bands: optical phenomenon from edge enhancement of enhanced darkness and lightness at borders, due to lateral inhibition of the retina. This is an inbuilt edge enhancement mechanism of the retina, where the edges of darker objects next to lighter objects will appear darker and vice versa, creating a false shadow. If you have high acuity, you don't detect change, small receptive fields. Enhancing contrast. Darker areas → lower firing rate in inhibitory ressponse bipolar cells→ inhibits sidebar less, appears lighter next to darker bar. Enhancing contrast has survival value. Less in-dark. Contrast and edge facilitation in stimulus: turn it into a response over space graph. Horizontal cells get input from photoreceptors. Stimulation from dark area, not light. Sends a lot of inhibitory output, more for the center. It mainly affects responses from sides. Inhibits more the sides, so they become less active. Photoreceptors stimulate horizontal cells, which inhibit photoreceptors and enhance contrast.. Edges get little stimulated because of their position, and because they are inhibited in the center you see even less in edges. Negative feedback loop: Photoreceptors stimulate horizontal cells, which inhibit photoreceptors and enhance contrast. The ones closer to the darker bar inhibit more. Middle bar: Horizontal cells in middle. They get differential input which receives different output. Half of the middle is in the dark, so they send a lot of inhibition back. The right side of the edge is associated with higher activity, becomes more inhibited so you see more dark next to the dark bar. Enhances contrast. 2. Bipolar cells: in retina. one bipolar cell connects to multiple receptor cells. You have multiple input, so you need less input to detect, lower threshold. Photoreceptor feed onto both horizontal cells as well as bipolar cells. Modulation by the horizontals affects the photoreceptors over time while they simultaneously stimulate the bipolar cells. The bipolar cells transfer the signals to the output stage of the retina: the ganglion cells. But they also modulate the signals. they respond to glutamate in 2 different ways because they have different types of glutamate receptors : Inverting: change the sign of incoming signals→ respond to light! Likes light surrounded by darkness. On-center off surround receptive field. on/inverting glu inhibits the cell, so responds to light. Invert their signal by using another type of neurotransmitter, they use glutamate. In mGluR6 receptor, glutamate is inhibitory. Hyperpolarized in dark. Excited with lack of neurotransmitter in light, no inhibition. Non-inverting: no sign change→ respond to darkness. Likes darkness surrounded by light. off/non-inverting glutamate excites cell, so it responds in darkness Ampa receptor → glutamate is excitatory. 3. Retinal ganglion cells Convergence: multiple inputs resulting in a single output. High convergence: no difference in binary response coming from different photoreceptors. Throw away spatial information differences, no discrimination between areas. But multiple cells converge, so they need very little input to respond because it is combined so the detection threshold goes down while discrimination goes up. Low convergence: light comes from a specific location to detect, harder. How the bipolar cell connects to the ganglion cells gives rise to a new type of receptive field profile: The center-surround receptive field. The center and surround respond to different inputs and must both be active. So the input to the ganglion cell depends on the local change. This special type of receptive field will be seen many more times in the visual system. Arrangement can have either sign, allow on-center (with positive from inverting cells) or off-center (positive from non-inverting cells) Center-Surround receptive fields of ganglion cells: respond to contrast contrast. Oncenter (light): If light is in the right place (center), there is a small response. If it is in the surround, inhibitory. If it fits, it is perfect, it fits the organization of the receptive field. If it fits surrounds, it sends more inhibitory feedback. Offcenter (darkness): other way around. Photoreceptors (dark)→ bipolar cells (light/dark) → ganglion cells (contrast) Mach bands at RGC level: Some darkness on and off surround, more active. This is why left edge is light, from the activity of on center off surrounding cells that spill over to the left side. So we can explain this at both the horizontal cell level but also at the retinal ganglion cell level. The ganglion cell have center-surround (CS) receptive fields (RF) and here it is all about receptive field fit. At the edges, the CS-RFs fit best, they have a difference between the center and surround. However, in the center (upper left in the example) of a mach band, there is no difference between the stimulation of the center and surround. So output is weak. In the light / dark cell: OnCenter-OffSurround more active, makes it look Brighter at the edge! In the dark /light cell (inverse): OffCenter-OnSurround more active, makes it look Darker at the edge! Hermann grid: Traditional explanation: at the intersections, receptive fields do not fit with surroundings. This difference in activity is due to the fact that at the intersections more surround inhibition is produced in ON-center retinal ganglion cells than at other sites. The same considerations apply to OFF-center ganglion cells when contrast is reversed. The fact that in the center of gaze the smudges are not perceived was attributedto the much smaller size of receptive fields in the foveal representation ○ In B, between 2, they respond pretty well because most of the surround is covered. But in the 4 edge intersection, they dont fit as well. Less darkness in the surround. However, orientation also matters. If orientation didn’t matter, the 2 stimuli should be the same in terms of illusory spots. But also eccentricity increases far away from the fovea, you dont see that well anymore because of less acuity. In the periphery, you only see a couple of black spots. And other reasons: The illusion is perceived over a large range of size, The illusion is reduced when the grid is rotated by 45, The illusion can be reduced or eliminated by manipulations that do not alter theantagonistic center/surrou nd activation of retinal ganglion cells. But it doesn’t work in other grids because receptive fields need to be elongated, not round. Alternative theory: Effect due to eye tremor, the illusory effect is just aspronounced, if not more so, under br ief exposures. This suggests that the effects maybe attributed to simultaneous processing of images. But mainly due to cells selective for oriented line segments may be involved in giving rise to the illusion. It is well known that the majority of cells i n area V1 are orientation-specific, the perception of lightness and darkness is the product of the relative activity of neurons driven by the ON and by the OFF systems, overall is product of both. propose that the illusory smudges are the result of the relative degree of activityof the ON and OFF S1 cells at the intersections, as compared with activity at non-intersecting locations They also have color selectivity. Tuning and things cells need to fire Cells with a center-surround receptive field prefer a spatial frequency, reflected in a certain size stimulus. This refers to the frequency of light/dark changes in space (waveforms). Polarity: on-center off surround needs to be correct Position in space Spatial frequency describes the periodic distributions of light and dark in an image. High spatial frequencies correspond to features such as sharp edges and fine details, whereas low spatial frequencies correspond to features such as global shape. Contrast sensitivity function: The sensitivity is related to the convergence in the retina (spatial frequency). Low frequencies have more convergence, leaving them more sensitive for detection so they need less contrast to be seen. However, if you go low enough, you simply don’t have the receptors anymore to perceive it (Arousal and Attention can both change the shape of this curve). We can only see medium spatial frequency. Depends on distance from the screen which changes receptive field size. Cats stare intensely in space: they can see lower spatial frequency, shapes that are not visible to us. Shadows from clouds. Sensitivity is related to spatial frequency: this is just a consequence of the functional anatomy (receptor distributions and densities in combination with preferred stimuli (size).