IPCV 2024/2025
Lecture 1 : Introduction
A digital image is a representation of a 2D image as a finite set of digital values called
picture elements or pixels. These pixel values typically represent gray levels, colors,
heights and opacities.
There are a total of four common image formats.
- 1 sample per point (B&W)
- 1 sample per point(Grayscale)
- 3 samples per point(Red, Green, and Blue)
- 4 samples per point(Red, Green, Blue, and ‘Alpha’ (opacity))
An image is in theory a function, from R2 to R.
- F(x,y) gives the intensity at a certain position (x,y)
o Expecting the image to be over a rectangle and finite
o 0 (intensity) = black
o 255 (intensity) = white
- RGB results in a vector format of these functions, three functions together.
- A digitization of an image implies that a digital image is a approximation of reality
, IPCV 2024/2025
Image Sources
A sources of acquiring images could be the electromagnetic energy spectrum
(sinusoidal waves with various wave lengths), as well as sound waves (ultrasound) or
electrons (electron microscopy).
The principal source for digital images is the EM energy spectrum.
Digital Image Processing
Image processing focusses on the improvement of pictorial information for human
interpretation. Processing of image data for storage, transmission and representation
for autonomous machine perception.
This process can be divided into a total of three processing levels.
Stages of Image Processing
, IPCV 2024/2025
Lecture 2: PSF, Convolution & Fourier
Point Sources
A surface with a light emitting point source can be described by the Dirac function. The
Dirac function is always 0, except when the two inputs are both 0, than the Dirac
function is equal to 1.
The same holds for a surface with two point sources.
When applying the same for multiple point sources on an orthogonal grid, the same
holds for the Dirac function. Delta represents the distance between these point
sources.
An infinite number of point souces on a grid can be visualized
by taking the limit from delta to 0, as an image can be regarded
as a sum of an infinite number of point sources arranged on a
grid.
, IPCV 2024/2025
Point Spread Function
The ideal case for image processing would be to only have point sources. However, due
to there being different lenses, these point souces ‘spread’. In this case, the dirac
function is not applicable, one should use the Point Spread Function (h).
On the left of this bottom image, a Dirac was used, resulting in points. On the right a PSF
was used to determine the resulting image. Using a PSF can be useful, resulting in a
more blurring image (lower resolution), however more details are revealed as to what is
being imaged as there are overlapping points.
Lecture 1 : Introduction
A digital image is a representation of a 2D image as a finite set of digital values called
picture elements or pixels. These pixel values typically represent gray levels, colors,
heights and opacities.
There are a total of four common image formats.
- 1 sample per point (B&W)
- 1 sample per point(Grayscale)
- 3 samples per point(Red, Green, and Blue)
- 4 samples per point(Red, Green, Blue, and ‘Alpha’ (opacity))
An image is in theory a function, from R2 to R.
- F(x,y) gives the intensity at a certain position (x,y)
o Expecting the image to be over a rectangle and finite
o 0 (intensity) = black
o 255 (intensity) = white
- RGB results in a vector format of these functions, three functions together.
- A digitization of an image implies that a digital image is a approximation of reality
, IPCV 2024/2025
Image Sources
A sources of acquiring images could be the electromagnetic energy spectrum
(sinusoidal waves with various wave lengths), as well as sound waves (ultrasound) or
electrons (electron microscopy).
The principal source for digital images is the EM energy spectrum.
Digital Image Processing
Image processing focusses on the improvement of pictorial information for human
interpretation. Processing of image data for storage, transmission and representation
for autonomous machine perception.
This process can be divided into a total of three processing levels.
Stages of Image Processing
, IPCV 2024/2025
Lecture 2: PSF, Convolution & Fourier
Point Sources
A surface with a light emitting point source can be described by the Dirac function. The
Dirac function is always 0, except when the two inputs are both 0, than the Dirac
function is equal to 1.
The same holds for a surface with two point sources.
When applying the same for multiple point sources on an orthogonal grid, the same
holds for the Dirac function. Delta represents the distance between these point
sources.
An infinite number of point souces on a grid can be visualized
by taking the limit from delta to 0, as an image can be regarded
as a sum of an infinite number of point sources arranged on a
grid.
, IPCV 2024/2025
Point Spread Function
The ideal case for image processing would be to only have point sources. However, due
to there being different lenses, these point souces ‘spread’. In this case, the dirac
function is not applicable, one should use the Point Spread Function (h).
On the left of this bottom image, a Dirac was used, resulting in points. On the right a PSF
was used to determine the resulting image. Using a PSF can be useful, resulting in a
more blurring image (lower resolution), however more details are revealed as to what is
being imaged as there are overlapping points.
