Introduction

Question 1

What is upsampling/downsampling and how is it done?

Answer 1

Upsampling is to increase the size of the image without changing its content.
Downsampling is the opposite. Its done by analyizing nearby pixels (for every pixel) and using these values to calculate new pixel values.

Question 2

What is it called when an object is being blocked by another object?

Answer 2

Occlusion

Question 3

Explain how low and high pass filters work.

Answer 3

Low pass filter blocks the highest signal values, or in the case of an image it decreases the disparity between nearby pixels which smooths the image.

High pass filter blocks the lowest signal values, and in the case of an image it sharpens the image. They both work the same, but use different kernels.

Question 4

Mention two areas where Computer Vision can be applied.

Answer 4

Safety measures
Health
Security
Autonomous anything (robots, self-driving cars etc.)
Basically anywhere...

Question 5

What is aperture size, and what happens when you adjust it?

Answer 5

Aperture size is the size of the “blockade” that a pinhole camera or similar creates. A bigger pinhole/aperture will let more light in, and as such it determines how much light hits the sensor. Generally, a wider aperture will give you a more shallow depth of field, whereas a smaller aperture will give you a deep depth of field.

Question 6

Mention at least two ways to enhance an image.

Answer 6

Remove noise
Increase contrast
Smoothing (using low pass filter)
Sharpening (using high pass filters)
Histogram manipulation

Question 7

How does a pinhole camera work?

Answer 7

Normally a sensor would get too much light from every direction to get a clear image. The pinhole camera focuses the light into a small area. This also causes the image to be turned upside down. A good example is simply the human pupil.

Question 8

What is a histogram, and how can it be used?

Answer 8

A histogram shows the number of all pixel values in an image (ranging from 0-255).

It can be used to:

Measure several things like averages, variance and contrast
Threshold selection
Image enhancement through equalization

Question 9

What is histogram equalization?

Answer 9

When you use the histogram to enhance the image. Adjusting contrast by modifying the intensity distribution

Essentially you make the pixel values more linearly distributed

Question 10

What is the general pipeline for image processing?

Answer 10

Localization: Where in the image is the object?
Object detection: Roughly separate the objects
Segmentation: Separate an object from background
Classification: Done after segmentation. What is this object?