W10 - SFA Channels, Applications & Monocular Cues Flashcards
How many cells underpin the contrast sensitivity function?
Is it due to having one cell type or multiple cells optimally tuned to different spatial frequencies? (single cell type vs. different population responses)
How would you set up an adaptation experiment to test this?
- Use electrodes in V1 to record different cells to look at different tuning
Behavioural studies: psychophysical electrode: use adaptation study
In conducting an adaptation study to determine the SF tuning of cells that
mediate the CSF, what does the participant view for an extended period of time
and what are they tested on?
Select a spatial frequency (eg 12 cycles per degree) and adapt to that stimulus - In adaptation CSF experiment, participants adapt to a pure sinewave
What pattern of results would indicate that the CSF is due to a single SF-tuned
channel?
all the different spatial frequencies will be reduced/less sensitive, and see a global overall reduction in sensitivity
What pattern of results would indicate that the CSF is due to multiple channels
tuned to different SFs?
Notch adaption =
the sensitivity of cells adapted to a specific spatial frequency will be reduced, while other cells that were not adapted to the spatial frequency will show no change in sensitivity, you only get a reduction in sensitivity around the cells that adapted to the stimulus with a specific frequency
Adoption study: 4. Which pattern of results do you get in those studies?
Get notch adaption (CSF is due to multiple channels
tuned to different SFs)
High SFs are processed by cells with small or large RFs?
High SFs are processed by cells with small receptive field sizes
Low SFs are processed by cells with small or large RFs?
Low SFs are processed by cells with large receptive field sizes
Is the image of Einstein defined by low or high SFs?
Einstein is going to have a lot of high spatial frequencies compared to Monroe
Is the image of Monroe defined by low or high SFs?
lower SFs compared to Einstein
In the Einstein-Monroe illusion - What happens to the SF of the image as viewing distance is increased?
/ Given the above, why do we perceive Einstein at short viewing distances,
and Monroe at long viewing distances?
- The spatial frequency content of the image gets higher with increasing viewing distance, and the more spatial frequencies are processed with smaller retinal image/visual angle and increasing distance
- By a certain distance, the high SFs content of Einstein becomes too fine to resolve under a human CSF, and thus the details of Monroe become more evident
Why does the illusion also work if you squint your eyes?
The process of squinting or looking from further away causes low pass filtering =
strips out the high spatial frequencies and the image is defined by low spatial frequencies
reduces your spatial acuity of the image, which thus reduces the sensitivity of high spatial frequencies to be processed.
You are more likely to make out the image of Monroe which is characterised by lower spatial frequencies
The painting of Mona Lisa has an ambiguous smile. We don’t see the smile when looking directly at it, only when looking away, i.e., in peripheral viewing.
(Explain this using SFA)
Since peripheral vision has less visual input to discriminate the finer details of the face, may lead to the fine details being average out with the shadows of the face, leading to blurring discrepancy of the face shadows and shape of the lips, less CSF components might appear to tell apart the shadows/lips = looking like she’s smiling
Answer = There are more HSF in fovea = looks like she is not smiling, and more LSF in periphery = looks like she is smiling
What is the idea of coarse-to-fine processing and how does is this consistent with the properties of magnocellular and parvocellular cells?
The visual system first processes low spatial frequency information and then fills in the details with high spatial frequencies
The axon conduction speed of p and m cells helps support the idea of course-to-fine processing in the visual system.
As Low SF processed by magnocellular system have a faster axon conduction speed than High SF processed by parvo cells, resulting in the visual system first interpreting an image by its course/outline content and then later processing the finer details by the p-cells.
What information is contained in a shadow boundary, compared to the
luminance boundary produced by a real object boundary?
How can the visual system can use the cell response to determine it is a shadow rather than a real object boundary?
The shadow border and real border both have low SF content
BUT
- the real border can be considered as a square wave of luminance, (characterised by both fundamentals and odd harmonics), so the real border has a COMBINATION of LOW and HIGH spatial frequencies to process the sharp edge (square wave) of the real border
- In the SHADOW border, all the cells tuned to LOW and INTERMEDIATE SFs but there are NO HIGH SF tuned to processing the edge
- Without HIGH SFs being tuned to the shadow, the visual system can determine that there is a discontinuity in luminance in the shadow border and processes it as a shadow instead of a real border / harsh contour.
What is the implication of using cell response to perceive shadows on harsh borders even when the object is gone?
Thus even though the object is no longer shown on the checkerboard, the brain still processes the luminance border as a shadow based on its spatial frequency content and its comparison of other spatial frequency content of the real borders of the checkerboard.
When watching a movie of a rotating wheel, what is aliasing? Explain why it occurs
Aliasing occurs when you misperceive the direction of motion in a stimulus due to the temporal frequencies shown are too high for the resolution of the visual system
(rotating more than 180 but less than 360 degrees, the smallest phase offset becomes in the opposite direction)
Aliasing occurs when the phase offset of a rotating object is HIGHER that the SAMPLING RATE of the RETINA from one frame of motion to the next
What do you need to do in order to avoid aliasing?
NYQUIST LIMIT = You need to sample at TWICE THE FREQUENCY that you want to resolve, eg. need
1. to resolve 10 hertz = need to sample at 20 hertz
In terms of input from the eyes, what are the two types of depth cues?
monocular and binocular depth cues
What are the two types of depth information?
- Absolute depth = gives you ACTUAL DISTANCE to the object
- Relative depth = the relative depth ordering between different objects - but absolute depth might not be known)
What are the two types of monocular depth cues?
- Pictorial = pictures, eg Examples of pictorial cues include pictures with:
Linear perspective
Texture gradient
Occlusion
Relative size
Aerial perspective
Elevation
Familiar size
Highlights and shading - Non-pictorial = not pictures
What are the seven pictorial depth clues and definition?
- Relative size between objects, if one object appears further away than another but they produce the same size on the retina, the brain upscales the further away object to generate an estimate larger perception
- Occlusion/interposition cue = the object that occulates another object will be perceived as closer in viewing distance
- Shadow = deeper shadows indicate greater depth
- Lighting = LESS depth is perceived when light is directed strongly onto a surface
- Clarity and elevation = Since clarity decreases with distance and light scatter results in reduced contrast, CLEARER OBJECTS ARE PERCEIVED AS CLOSER. With higher elevation, objects are perceived as further away, eg. birds, aeroplanes
- Linear Perspective = objects in depth recede to a single point, so the lines that are physically parallel produce 2D images in which they converge
- Texture gradient = With increasing distance, the spatial frequency content of the image goes towards high spatial frequencies (eg. Monroe/Einstein illusion, more dense texture/detail), higher texture can give an impression of depth
How can you make advertising signs that are drawn onto a on a field appear to stand vertically up?
The image is drawn as an elongated rhombic shape ,which appears more flattened/protracted with increasing distance, such that the visual system will perceived it as a square at the visual angle that it is viewed from
(From the opposite direction, the distortion and the perspective distortion act in the same direction and the elongated shape will look even more elongated!)
What happens to the SF spectrum of texture gradient as distance to it is
increased?
With increasing distance, the spatial frequency content of the image goes towards high spatial frequencies (eg. Monroe/Einstein illusion) and the SF spectrum will be higher / more towards the right of the x-axis plotting contrast over SF
What is accommodation?
How does it change the shape of the lens at different distances?
Accommodation is the automatic process of changing the shape and thus power of the optic lens to be able to focus on objects at different viewing distances and maintain clear vision
- At short distance = lens is more spherical and more refracted to bend at close incoming light
- At long distances = lens is more flatter and less refracted
How can accommodation be used as a depth cue?
- If we could SENSE the accommodative state for our lens = it could be cue for ABSOLUTE DEPTH AND DISTANCE, but only chameleons can do it
Humans can’t use it as a depth cue, but what experiment was conducted to show that chameleons can use it?
- Give the chameleon glasses with increased refractive power to change the accommodative state of the chameleons’ eyes
- Test whether the chameleon’s ability to use accomodation for absolute depth has been impact
- Results = With distorted accommodation, the chameleon underestimates the distance to the bug
- increased refractive power, chameleon uses less accomodation and assumes the bug is further away, puts tongue out too long
What are the two kinetic cues to depth?
- Motion parallax = the motion of the OBSERVE results in relative motion depth
- Everything in front of you moves in the OPPOSITE DIRECTION TO YOU
- THINGS UP CLOSE MOVE REALLY QUICKLY, THINGS FARAWAY MORE SLOWER
- Kinetic depth effect (KDE) = DYNAMIC MOTION CUE, produced by object motion with DIFFERENTIAL VELOCITIES, SPEED AND DIRECTION by a MOVING object, to provide clues to the 3D shape of an object
EXAMPLE = We can observe differential changes in speed, velocity and direction to make inferences about the mood and gender of a stick figure against a black background