Image Processing and Human Vision Quiz

Course Information

• Course Title: Introduction to Computer Vision
• Course Code: CPS834/CPS8307
• Instructor: Dr. Omar Falou
• University: Toronto Metropolitan University
• Semester: Fall 2024

Digital Image

• Digital Image: A two-dimensional function where x and y are spatial coordinates. The amplitude of f at the point (x,y) is called intensity or gray level (f(x,y)).
• Pixel: The fundamental elements of a digital image.

Image Sources

• Electromagnetic (EM) energy spectrum (Gamma rays, X-rays, ultraviolet, visible, infrared, microwaves, radio waves).
• Acoustic
• Ultrasonic
• Electronic
• Synthetic images produced by a computer

Electromagnetic (EM) Spectrum Examples

• Gamma-ray imaging: Used for nuclear medicine, astronomical observations, microscopy, & biological imaging.
• X-rays: Used for medical diagnostics, industrial applications, & astronomy.
• Ultraviolet: Used for lithography, industrial and astronomical observations.
• Visible & infrared bands: Used for light microscopy, astronomy, remote sensing, industry, and law enforcement.
• Microwave band: Used for radar.
• Radio band: Used for medicine (MRI) and astronomy.

Examples of Imaging Techniques

• Gamma-ray imaging (Bone scan, PET image, Cygnus Loop, Gamma radiation)
• X-ray imaging (Chest X-ray, Aortic angiogram, Head CT, Circuit boards)
• Ultraviolet imaging(Normal corn, Smut corn, Cygnus Loop).
• Light microscopy imaging (Taxol, Cholesterol, Microprocessor, Nickel oxide, Surface of audio CD, Organic superconductor).
• Visual and Infrared imaging (LANDSAT satellite images of the Washington, D.C. area).

Automated Visual Inspection

• Examples of manufactured goods that are often checked using image processing (A circuit board controller, Packaged pills, Bottles, Air bubbles in plastic products, Cereal, Image of intraocular implant).

Automated Visual Inspection - Results

• Results of automated reading of the plate content by the system.
• The area in which the imaging system detected the plate.

Example of Radar Image

• Spaceborne radar image of mountains in Southeast Tibet.

Example of MRI

• MRI images of a human knee and spine.

Example of Ultrasound Imaging

• Examples of ultrasound imaging(Baby, Another view of baby, Thyroids, Muscle layers showing lesion).

Image Processing → Computer Vision

• High-level: Object detection, recognition, shape analysis, tracking, and use of artificial intelligence and machine learning.
• Image Analysis: Segmentation, image registration, and matching.
• Low-level: Image enhancement, noise removal, restoration, feature detection, and compression.

Image Processing Problems

• Image acquisition, Image enhancement, Image restoration, Morphological processing, Segmentation, Representation & description, Object recognition, Image compression.

Element of Visual Perception

• Although the digital image processing field is based on mathematical and probabilistic formulations, human intuition and analysis also play a central role. The choice of technique often depends on subjective visual judgments.
• Developing a basic understanding of human visual perception is essential.

Structure of the Human Eye

• Cornea: Tough, transparent tissue covering the anterior surface of the eye.
• Sclera: Opaque membrane enclosing the remainder of the optic globe.
• Choroid: Lies beneath the sclera, containing a network of blood vessels, the major source of nutrition to the eye.
• Lens: Absorbs infrared and ultraviolet, but excessive amounts can damage the eye.
• Retina: The inner membrane lining the posterior portion of the eye's wall, where light from an object outside the eye is imaged.

Receptors

• Receptors (neurons): distributed over the surface of the retina.
• Divided into cones and rods.
• Cones (neurons 6-7 million): Primarily located in the fovea (central portion), highly sensitive to color and details.
• Rods (75-150 million): Distributed over the retina, highly sensitive to dim light.

Blind Spot

• Figure 2.2 shows the density of rods and cones in a cross-section of the right eye, passing through the optic nerve's emergence point.
• Blind spot: absence of receptors, radially symmetrical around the fovea.

Image Formation in the Eye

• The eye's lens is flexible, unlike ordinary optical lenses; its focal length varies.

Brightness Adaptation and Discrimination

• Perceived brightness is not solely determined by intensity.
• Visual system tends to undershoot or overshoot around the boundary of regions with different intensities (Mach bands).
• Simultaneous contrast: a region's perceived brightness depends on the surrounding regions.

Optical Illusions

• Examples of well-known optical illusions.

Light and EM Spectrum

• Colors humans perceive are determined by the reflected light wavelengths from an object.
• Wavelengths for example: green roughly 500 to 570 nm.

Light and EM Spectrum – continued

• Monochromatic light: devoid of color. Intensity varies from black to white, resulting in grayscale images.
• Chromatic light bands: the range of wavelengths that create color.
•  Radiance (W, watts): the total energy emitted by a light source.
• Luminance (Im, lumens): the amount of perceived light by an observer.
•  Brightness: a subjective descriptor of light perception.

Image Acquisition

• Transforming light energy into digital images.
• Components: energy, filters, power source, housing, sensing material, output voltage waveform.
• Images can be acquired using single sensors or sensor strips (linear or circular).

Simple Image Formation Model

• Model of image formation: f(x,y) = i(x,y) * r(x,y).
• i(x,y): Illumination intensity at a point (x,y).
• r(x,y): Reflectance or transmissivity at (x,y) describing the object's light interaction.

Components of Image Formation Models

• f(x,y) : Intensity image at coordinates (x,y)
• i(x,y) : Illumination component at (x,y).
• r(x,y) : Reflectance component at (x,y), representing the object's light reflection.

Some Typical Illumination Ranges

• Lumen: a unit of light flow or luminous flux
•  Lumen per square meter (lm/m²): The metric unit used to quantify illuminance of a surface.
• Typical illumination levels: daytime (clear & cloudy), evening (moonlight), commercial office.

Some Typical Reflectance Ranges

• Different materials have different reflectance levels.

Gray Level

• Intensity of a monochrome image.
• Common gray scale is [Lmin, Lmax] (0-black, 255-white).

Image Sampling and Quantization

• Digitizing image coordinates and amplitude values: converting continuous into discrete values.
• Sampling: the division of the image into discrete samples across the x and y axis. Pixels.
• Quantization: dividing the range of light intensities values to assigned to each coordinate into discrete values. These discrete values are often represented by the number of bits assigned to the pixel.. 

Coordinate Convention

• Defining the origin (0,0) and orientation (x-axis, y-axis) of the coordinate system, typically positioned at top-left corner within a frame.

Representing Digital Images

• Representing images as a matrix where the integer values within the matrix correspond to pixel values (intensity).

Representing Digital Images – Continued

• Discrete intensity interval: [0, L-1] with L = 2k (where k is the bit depth).
• Stored number of bits, b, for an image of size M×N.
• Image with 256 possible gray levels, often referred to as 8-bit image.

Representing Digital Images – Table

• Table showing number of bits necessary for different image sizes and bit depths.

Coloured Image

• Represented by RGB channels.
• RGB: Red, Green, Blue

Spatial and Intensity Resolution

• Spatial resolution: smallest discerned detail in an image (e.g.; dots per inch).
• Intensity resolution: the smallest discernible change in an image intensity (e.g., bits).

Image Interpolation

• Process of estimating unknown image values using known values.
• Methods- Nearest Neighbor, Bilinear, Bicubic. 

Distance Measures

• Pixel distance metrics, such as Euclidean, City Block, and Chessboard distances, define ways to measure the proximity between pixels in a digital image.

Mathematical Operations in Digital Image Processing

• Array operations are distinct from matrix operations.
• Matrix product is a mathematical operation involving matrices.
• Array product involves the element-wise multiplication of corresponding elements in image arrays.

Linear vs. Nonlinear Operations

• Additivity: H(a1f1(x,y)+a2f2(x,y)) = a1H(f1(x,y)) + a2H(f2(x,y)).
• Homogeneity: H(af(x,y)) = aH(f(x,y))

Arithmetic Operations

• Basic operations are used for manipulating pixel values, using addition, subtraction, multiplication, and division.

Example: Addition of Noisy Images for Noise Reduction

• Method of reducing noise in images, such as those from astronomical observations, by averaging multiple noisy images.

Example: Image Subtraction: Mask Mode Radiography

• Method of enhancing contrast in images, such as X-ray images, by subtracting a mask image from a live image.

Example: Image Multiplication

• Images are multiplied together to change or enhance characteristics within the images, such as modifying highlighting or shading within an image.

Set and Logical Operations

• Set theory operations are used for pixel-based image manipulation.
• Set operations allow using logical operations to create new images by comparing pixel values from different image sets.

Set and Logical Operations – Continued

• The image values can be represented using set theory and the resulting sets can be used to perform a variety of image operations on a pixel-basis.

Spatial Operations

• Single-pixel operations and neighborhood operations. These affect pixel values based on either the intensity of a single pixel value, or pixel values from neighboring pixels. 

Geometric Spatial Transformations

• Transforming coordinate systems from one image to another, such as enlarging, shrinking, or rotating images.

Geometric Spatial Transformations – Continued

• Affine transform: A commonly used spatial transformation that encompasses scaling, rotation, and translation of images. 

Intensity Assignment (Forward and Inverse Mapping)

• Methods of assigning pixel values in output images based on their corresponding coordinates in input images to maintain spatial integrity during transformations such as rotation. 

Example: Image Rotation and Intensity Interpolation

• Different methods of assigning intensities to new locations in rotated images based upon nearest neighbor, bilinear, and bicubic interpolation.

Image Registration

• Methods for aligning images; determining how one image is positioned related to another; usually based on known correspondences between landmarks between images.

Image Registration – continued

• Tie points or control points are used to align (register) images. A bilinear approximation model: mathematical equation to map images (by finding a transformation function that models the relationship between images).

Image Transform

• Transforming input and output images for different image process applications. Forward and inverse transformation kernel functions can be used to transform between domains. The transform domain is often different than the spatial domain and may facilitate operations that are difficult or impossible to complete on direct image transformations.

Test your knowledge on image processing concepts and human visual perception. This quiz covers topics such as focal length, brightness perception, color detection, and image resolution. Perfect for students of optics and digital imaging.