Image Processing Fundaments Flashcards

1
Q

What does the MTF measure

A

The ability of an optical system (eye, camera) to reproduce (transfer) various levels of detail (spatial frequencies) from the object to the image.

How good is the lens at transmitting light when the light is distributed in a sine wave fashion? These sin waves can be low or high spatial frequency.

Y axis: Spatial frequency
X axis: Image contrast/object contrast (ratio 0-1)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Low vs high spatial frequency

A

Low- fat bars.

High- skinny bars.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

the optical system is better at transferring ___ spatial frequencies

A

Low. For low spatial frequencies, the image contrast is very close to the object contrast. Value near 1.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is different between the optical MTF and the neural MTF (RTF) *both log graphs

Which one is better able to transfer contrast?

A

The optical MTF determines the contrast transfer level of combined optical components of the eye.
-low pass in shape.

The neural MTF (RTF) determines the contrast transfer level of the neural retina with optics bypassed.
-Band pass in shape

Take away: Retina is better able to transfer contrast. Why? BC the graph is higher up- closer value to 1. This means that it has a high sensitivity= better detection

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What is the contrast sensitivity function?

A

Sum of components contributed by the optics (MTF) and neural retina enhancement (RTF)

CSF is perceptual. Ask its to tell us what they see.

X: Spatial frequency
Y: Contrast sensitivity and contrast ratio

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

The CSF demonstrates that human vision acts as a ___ spatial filter.

A

Bandpass

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Where CSF crosses the x axis

A

High spatial frequency cut off. Can convert to snellen acuity. Cuts off at 60 cpd, 20/20 vision.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

You would think that for fatter bars (low spatial frequency), you could see it without as much contrast. Why can’t we?

A

Because lateral inhibition kicks in when target is stationary. However if you flicker or move the low spatial frequency grating, you won’t ned as much contrast to detect it. This creates a low pass function instead of a band pass function.

Therefore, we are not as sensitive to detect contrast of super low or super high gratings. There is a high and low spatial frequency cut off.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Influence of retinal eccentricity upon the CSF

A

In order for testing to be fair, you must magnify the image presented in the periphery based on receptive field differences. If you normalize this, the CSF for peripheral and foveal locations will be the same.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Influence of aging upon the CSF

A

No change occurs for low spatial frequencies, but high spatial frequencies drop and shift to the left. Overall height of CSF drops.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Influence of optical blur upon the CSF?

A

Does not affect low spatial frequencies or the peak, but high spatial frequencies shift to the left. The high spatial frequency cut off is sooner, indicating worse vision.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Influence of glare upon the CSF?

A

Lowers entire CSF graph- low, peak, and high spatial frequencies are lower (less sensitivity).

Therefore, glare reduces CSF more so than blurry vision.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Influence of infants upon the CSF?

A

Infants overall ability to detect contrast is much worse than adults. A 1 month infant shows a low pass curve. high spatial frequency cut off is very low= worse vision.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Why do babies have a lower sensitivity?

A

Cone morphology: Cones don’t capture photons as well as adults.

Less cone density: Causes high spatial frequency cut off to occur sooner.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What is the spatio-temporal MTF

A

A complete description of human visual sensitivity. Has changes in contrast over both space and time. Allows 3D function.
X: Temporal frequency
Y: Detection threshold. Minimal contrast required to detect stimulus.
Z: Spatial frequency

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

“Notch” defects in contrast sensitivity could be due to which 3 things

A
  1. Under or over corrected refractive error. Related to pupil size and luminance level.
  2. Parkinsons disease. Orientation specific due to disruption of neurons.
  3. Paracentral scotomas in glaucoma.
17
Q

A single neurons receptive field can be thought of as a ___ spatial frequency filter

A

Band pass

18
Q

Spatial frequency channel

A

To cover the entire range of spatial frequencies human perceive, neurons with different spatial frequency tuning are needed.

All neurons with similar “spatial frequency tuning” that work together, but independently of other neurons with different spatial frequency tuning, constitute a spatial frequency channel.

19
Q
  1. High spatial frequency loss can be due to
  2. Mid/low spatial frequency loss can be due to
  3. Overall spatial frequency loss can be due to
  4. Notches
A
  1. Mild refractive error, visual deprivation, mild ambylopia
  2. MS
  3. Cataracts, severe refractive errors, severe amblyopia.
  4. Paracentral glaucoma
20
Q

Prolonged viewing of high contrast patterns (adaption) showed that it makes subsequently viewed similar patterns harder to see.

EX: Look at stimulus #1 for 5 minutes, then stimulus #2.

A) What does it mean if you need to increase contrast in stimulus 2 to detect it?

B) What does it mean if you didn’t need to increase contrast in stimulus 2 to detect it?

A

A) Means that stimulus 1 and 2 used the same spatial frequency channel.

B) Means that stimulus 1 and 2 used different spatial frequency channels.

21
Q

If we only had 1 neural channel, adapting to a single spatial frequency should ____ our sensitivity to all spatial frequencies.

If we have multiple neural channels, then adapting to a single spatial frequency should ___ our sensitivity only to a narrow range of similar spatial frequencies.

A

Decrease and decrease

Ex: Adapt to 7 cpd. Ability to detect other gratings was not affected, only similar gratings close to 7cpd.

22
Q

Adaptation to sinusoidal grating changes the perceived spatial frequency of other subsequently viewed gratings- an illusion.

Adapt to any pattern.

a) If you look at a higher SF than the adapted pattern, it will appear:
b) If you look at a lower SF than the adapted pattern, it will appear:

A

a) It will appear to have an even higher SF than it actually does. Skinnier bars will be perceived.
b) It will appear to have an even lower SF than it actually does. Fatter bars will be perceived.

This causes a distribution shift of perceived spatial frequencies in the brain.

Ex: You adapt to a high SF. Causes your high SF neurons to fire a lot and not be able to increase their response. When you change the stimulus to a lower SF, high SF neurons must recover and low SF are able to fire. Causes you to perceive stimuli even more lower than they actually are.

23
Q

evidence of SF channels (4 examples)

A
  1. Spatial frequency (size) adaptations. Adapt to a pattern. If you look at a stimulus with higher SF than adapted pattern, the stimulus will appear skinner than it is. Look at a stimulus with lower SF than adapted pattern, the stimulus will appear fatter than it actually is.
  2. Size induction (Same time. Look at 2 same size center circles surrounded by either small or larger circles. The center circle surrounded by small circles will look larger.
  3. Spatial frequency masking. Masking only occurs when 2 patterns have very similar spatial frequencies. Checker board example.
  4. Sub-threshold summation. 2 sub threshold stimuli can be added together to reach threshold if they are using the same channel, or still be undetected if they are using different channels and cancel each other out.
24
Q

Evidence of SF channel- Size (spatial frequency) induction

A

Same time. Ex: Look at center circle, which looks larger? Both same size, but one looks larger to due multiple neural channels.

25
Q

Evidence of SF channel- SF masking

A

Masking only occurs when 2 patterns have very similar spatial frequencies, not dissimilar ones.

Ex: look at checker board. Since sharp edges are transmitted by the same system that transmits high SF and there are a lot of sharp edges, the high SF info masks the low SF info. Remove the mask by applying a low pass spatial filter. Now low SF will be easier to see.

26
Q

Sharp edges are transmitted by the same system that transmits __ SF

A

high

27
Q

Evidence of SF channel- Sub threshold summation

A

2 gratings each whose contrast is below a viewers threshold. If added together, may exceed the viewers threshold and become visible. This occurs only if their spatial frequencies are similar enough for them to share the same channel.

If the two subtreshold gratings each stimulate different channels, then their sum should NOT be visible.

28
Q
  1. Large neural receptive fields are better at detecting ___ SF
  2. Small neural receptive fields are better at detecting ___ SF.
A
  1. Low.

2. High.

29
Q
  1. Low SF filters encode:

2. High SF filters encode:

A
  1. Low SF encodes coarse luminance- luminance is slowly changing. large objects, overall shapes.
  2. High SF encodes fine spatial structures. Small objects, edges, detail.