Week 6 (T15): VR, AR, MR and Subject Review Flashcards
What distinguishes virtual reality (VR) from augmented reality (AR) and mixed reality (MR)?
Virtual Reality (VR) immerses users entirely in a computer-generated environment, replacing the real world. In contrast, Augmented Reality (AR) overlays digital information onto the real world, enhancing the user’s perception of reality. Mixed Reality (MR) blends both real-world and virtual elements, allowing interaction with digital content while being aware of the physical environment.
Image Acquisition, Sampling, and Quantisation
- Digitising analog signals using sensors
- Sampling reduces continuous data to
discrete values - Quantisation assigns discrete values to
ranges
Colour Spaces
RGB: Red, Green, Blue
HSV: Hue, Saturation, Value
YIQ: Luminance, In-phase, Quadrature
Image Content Representation
- Pixels: smallest unit of an image
- Features: characteristics of an image (e.g.,
edges, corners) - Descriptors: numerical representations of
features
Digital Images vs. Real World
- Digital images are discrete approximations
of the real world - Properties include width, height, bit depth,
etc.
Filtering Operations
- Apply mathematical functions to modify
pixel values - Examples include smoothing, sharpening,
edge detection
What are the four steps involved in image acquisition?
- Focusing light onto a sensor
- Converting light intensity to electrical signals
- Sampling the continuous signal to discrete values
- Quantizing the discrete values to a specific range
Explain the difference between RGB and HSV color spaces.
RGB is an additive color space where primary colors (red, green, blue) are combined to create other colors. HSV is a subtractive color space where hue represents the color itself, saturation indicates the intensity of the color, and value represents brightness.
What are the advantages of using feature descriptors compared to raw pixel values?
Feature descriptors are more compact and robust to image variations like noise and illumination changes, making them better for image recognition and retrieval.
Why are digital images considered discrete approximations of the real world?
Digital images represent the continuous information of the real world through a limited number of pixels and quantization levels, leading to loss of information.
Describe the purpose and effect of filtering operations on images.
Filtering operations modify pixel values based on mathematical functions to achieve specific effects like smoothing, sharpening, edge detection, etc. These operations can enhance specific image features for analysis.
Compare and contrast the early fusion and 3D CNN approaches for video classification.
Both use multiple video frames as input, but 3D CNNs also exploit temporal information through kernels that slide along the temporal axis. This allows them to learn motion information and can be more resource-intensive than early fusion.
Explain the principle of fairness in responsible AI.
AI systems should treat all people fairly and avoid biased predictions based on factors like race or gender that are unrelated to the task at hand.
What are the potential risks associated with data exposure in AI applications?
Sensitive data used to train or operate AI models, such as medical records, could be exposed in a data breach and cause harm to individuals.
An image has 1024x768 pixels and uses 256 quantization levels per color channel. How much memory does it require?
3 bytes/pixel * 1024 pixels * 768 pixels = 2,359,296 bytes
