Please enable JavaScript.
Coggle requires JavaScript to display documents.
Fundamentals of Visual Computing - Coggle Diagram
Fundamentals of Visual Computing
The Fourier Transform
Analogue signal: A continuous signal that contains time-varying quantities
Fourier Series: Any periodic function can be represented by a weighted sum of sines and cosines of different frequencies
Complex numbers
Euler's formula: Establishes the fundamental relationship between the trigonometric functions and the complex exponential function.
Properties
Linearity
Shifting
Modulation
Scaling
Convolution
Geometric Transformations
Images and Colours
The visible light spectrum is 380-750nm
Speed of light in a vacuum = frequency * wavelength
Digital images are 2D sampled representations of some continuous function, like a real scene
Smallest individual element of an image is a pixel which is a single sampled colour represented by 8-bit RGB (256 colours)
Rasterization: convert a continuous vector image representation to a rectangular sampled grid of pixels.
Image storage
Greyscale: arranged in scanline order, left to right & top to bottom
Colour: arranged in memory with rows of R,G,B pixels
You would access pixel x,y at 3
Cy
w[width of pixel] + x
Image compression: lempel-ziv-welch lossless coding method or DCT (discrete cosine transform) used in JPEG which is lossy.
Colour Spaces
RGB - Additive: used for monitors as the LED colours mix
CMY - subtractive: used for printing
CMYK usually used in printers because black is most common colour used and without it being separate, CMY would all have to be used which would not give an accurate black colour and would be more expensive.
HSV/HSL (Hue, Saturation, Lightness/Value)
Image sensors
CMOS: Each pixel contains an amplifier so it can be read out faster
CCD: Serial devices where pixels are read out one at a time.
Imaging. Put an aperture between the sensor and the object so that each pixel receives one ray of light. a lens can also help to direct and focus the light rays.
A histogram is a graph showing the distribution of intensities in an image
Filtering
Linear filtering: Slide a filter (kernel) over the image to compute a new value for each pixel. (Convolution)
Gradient filters (sobel filters)
Motion blur
Sharpen
Blurring. Can use box filter where each element of the kernel has equal weights and is normalised or can use gaussian blur which is smoother and preserves edges better
Median Filtering: replace each pixel with the median value of all pixels in its neighborhood. good for removing noise but you lose detail.
Morphological filtering: Filtering of image using set theory.
Geometric Modelling
Types of transform
Rigid (most common)
Articulated (piecewise rigid)
Non-rigid (Very complex)
Can transform a shape using matrix operations
Types of matrix transformation
Scale
Stretch
Flip
shear/skew
rotation
Add a 1 to the vector to make it homogeneous