Basic Concepts of Computer Vision — part 3
Some of the basic concepts that are very popular in computer vision are briefly explained in these computer vision series of articles. In this part 3 following topics are discussed:
- Digitization
- Aliasing
- Digitization of images
- Dynamic Range
- Color Interpolation
- CFA Interpolation
- Color Correction
- Gamma Correction
- Color Constancy
- White Balance
- Automatic White Balance
- Gray — World Assumption
- Automatic Focusing
- Automatic Exposure
- Resolution
Digitization
Digitization involves converting continuous signal of visual information into discrete pixels or quantised samples.
Aliasing
If we miss any signal in the sample, we can do Aliasing. A continuous signal can be reconstructed from its samples, providing the sample rate is correct.
Digitization of images
Digitization of images is the process of converting analog visual information into digital image files. For example capturing images with digital cameras and storing them in a digital format like png .
Dynamic Range
The range of light levels that a camera or imaging system can produce or capture. It is the difference between brightest and darkest areas an imaging device can show.
Color Interpolation
A technique used to estimate the colour values of pixels in a digital image when colour information is not available for every pixel. There are many ways of doing this, may including reconstructing RGB colour samples for each pixel from Bayes colour sample. Another way can be replicating the corresponding values of nearest neighbor pixel as colour doesn’t change significantly in neighbours.
CFA Interpolation
CFA (Colour Filter Array) interpolation is a specific form of colour interpolation used in digital cameras that have a sensor with a CFA, such as a Bayer filter. It interpolates missing colour samples based average of the neighbour pixels.
Colour Correction
The process of adjusting colours in an image to achieve desirable colour reproduction is called as colour correction. Colour correction digitally compensates the spectral sensitivities of CFA to bring them closer to CIE colour matching functions.
Gamma Correction
It is a nonlinear adjustment applied to the intensity values of an image to compensate for the nonlinear relationship between pixel values and perceived brightness. It ensures that the displayed image appears more natural to the human eye.
Color Constancy
Colour constancy means if two sources of light have different spectrums, a white object should change its colour. However the perceived colour of object remains relatively constant under varying illumination conditions.
White Balance
White balance is the adjustment of an image’s colours to make the white areas appear neutral and free from unwanted colour casts. It helps to keep the image in its true colour no matter what different lightening conditions were used.
Automatic White Balance
Every digital camera is equipped with a processing module that provides colour constancy this feature is known as Automatic white balancing.
Gray — World Assumption
Gray world assumption means if an image has sufficient amount of colour variations, the average value of the Red, Green, and blue channels should be same or average out to common gray value.
Automatic Focusing
Auto focusing means tuning of relevant parameters of the lens to maximise image sharpness.
Automatic Exposure
It is a camera feature that adjusts the exposure settings (shutter speed, Illuminance) to ensure that the image is properly exposed, neither too bright nor too dark.
Resolution
Resolution is amount of detail a camera can capture. It depends on, sensor resolution, optical resolution, sharpness of focus, and CFA interpolation.
References
E. R Davies, Computer Vision: Principles, Algorithms, Applications, Learning, Academic Press; 5th edition; 2017
D Forsyth, Computer Vision. A Modern Approach
