Image Representation Flashcards
When designing a computer system what do we use as an indication that a visual problem is solvable?
If a human can solve it
What is the purpose of the retina?
Encode the nerve pattern into the nerve signals
How can we see objects at different ranges?
Changing the refractive power of the eye’s lense
Why does the iris dilate in dark conditions?
To enable us too see clearly in dark places
What does the iris do?
Regulate amount of light entering eye
What are spatial differences in an image?
Difference between data in nearby regions
What computer vision concept is spatial difference related to?
Edge detection
What are temporal differences?
Differences in the specific image data over ti,e
How is RGB related to the human eye?
Cones (retina cells) red green or blue light sensitive
What camera is used to capture most Computer Vision images?
Video
Is the image we derive from a camera accurate?
No, inaccuracies in process
- radiant energy
- focus onto photosensitive detector
- absorbed
- measured digitally
- assemble measurements into image
- MEASUREMENT UNCERTAINTY ASSOCIATED WITH EACH PIXEL VALUE
How is the uncertainty in an image quantified?
Signal/Noise
Sources of degradation within camera?
- lens
- detector (suboptimal responce)
What is the grey value (0-255) of pixels in overexposed regions of photographs?
255
What do we define as noise in an imaging system?
Any deviation of the SIGNAL from its EXPECTED VALUE
What are the two types of speckle effect seen?
Salt and pepper (uncommon)
Gaussian (common in all systems)
What does an image histogram describe?
frequency of each grey level ocurrence
What can the image histogram be used for?
a measdure of noise amplitude
What is signal/noise measured in?
Decibel
Equation for signal/noise?
SNR = 10* log(signal/noise) signal = 255(max)
When a vision system is being designed, is the resolution of the images to be processed relevant?
Yes, it is required
What did Nyquist observe with his sampling theorem?
“to represent a periodic signal unambiguously requires at least two samples per second”
How does the nyquist theorem translate to a spacial resolution requirement from a vision system?
“Two pixels must span the smallest dimension of an object in order for it to be seen in the image”
What is another way to specify the spacial resolution?
θ = r/Z
r: smallest resolvable object
θ: field of view
Z: range, object distance
