Tales of Science & Data
MetaAbout meGitHubTwitter
Search…
Tales of Science and Data
Meta & resources
The meta on all this
Beautiful web of data science
Probability, statistics and data analysis
Probability, its interpretation, and statistics
Foundational concepts on distribution and measures
Hypothesis testing
Methods, theorems & laws
Notable brain teasers, paradoxes and how to be careful with data
Machine Learning: concepts & procedures
Overview of the field
Learning algorithms
Feature building and modelling techniques
Dimensionality reduction and matrix factorisation
Machine Learning: fundamental algorithms
Learning paradigms
Supervised learning
Unsupervised learning
Machine Learning: model assessment
Generic problems models can have
Performance metrics and validation techniques
Diagnostics
Artificial neural networks
Overview of neural networks
Types of neurons and networks
Natural language processing
General concepts & tasks in NLP
Manipulating text and extracting information
Topic Modelling
Word Embeddings
Computer vision
Intro: quantifying images & some glossary
Notes on the formalism
Processing an image
What's in an image
The computer science appendix
What's this
Notes on foundations
Essential algorithms
The mathematics appendix
Matrix algebra notes
Mathematical functions
Some geometry
Cross-field concepts
(Some) mathematical measures
Toolbox
The Python data stack
Databases and distributed frameworks
Notebook tools
Powered By GitBook
Intro: quantifying images & some glossary
Under the name "Computer vision" (CV) goes that set of techniques which allow us to manipulate and extract numerical data, and higher-level information, from an image, via a computer. This section will explain the basics of the field and then deal into more details, outlining the concepts that CV tackles. Nowadays, there is a division in the field in regards to "traditional" CV (the set of techniques which is not based on the use of Neural Networks/Deep Learning) and more state-of-the-art methods based on ANNs. We will touch on both sides.
Computer Vision is very interesting - it's about writing code to manipulate and use images and it allows for many applications. Have a read at the two blogs in the references for a bit on the relation between Computer Vision and Machine Learning/Deep Learning.
For the code here in this project, we will use OpenCV through its Python bindings. For the part regarding ANNs, we will make use of TensorFlow and Keras.
The glossary below is meant to contain basic definitions used all over the chapter.

Some glossary

Bimodal image

An image with two peaks in the intensity histogram.

Binary image

Is an image with pixels which can only take values 0 or 1.

Intensity

The intensity is the pixel value and is defined for grayscale images, where it is an 8-bit integer, hence taking values in the range [0:255], where 0 is black and 255 is white. It is sometimes referred to as brightness.

Colour

For coloured images, colour is given as a vector of 3 components of intensity. For example in the RGB values are the intensities for red, green and blue.

Contrast

Is the difference between maximum and minimum pixel intensities.

ROI (Region of Interest)

A part of image selected for a purpose, like cropping around it or cutting it out.

Formalism

Some notes on the formalism used are written in this page:

References

    1.
    A good and super brief outline of the history of the field and its relation to Machine Learning in T Malisiewicz's blog​
    2.
    An even better and more detailed outline of the development of Computer Vision and its relation to Deep Learning, from the same blog above
Previous
doc2vec
Next
Notes on the formalism
Last modified 6mo ago
Copy link
Contents
Some glossary
Bimodal image
Binary image
Intensity
Colour
Contrast
ROI (Region of Interest)
Formalism
References