MM1041 Chap.7 Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
Why vision matters
The e-learning lecture begins with people: about ninety per cent of the information sent to the brain is visual, and ninety-three per cent of human communication is said to happen through visual forms.
A machine that can interpret images therefore reaches a large share of the information businesses and people rely on.
Definitions and history
Early definitions describe computer vision as automatically identifying objects in images or videos; broader ones describe understanding the world through visual perception, and many people think of it as simulating the human eye.
The field began before the 1970s with digital image processing, added pattern recognition, and has recently moved to deep learning and to systems that combine vision with language. Its stated goal is to write programs that interpret an image and tell its story: what is present, where, and how things relate.
What a computer sees
A computer sees numbers.
An image is a grid of pixel intensity values; in an 8-bit image the brightest value is 255, and whiter pixels carry higher values. Everything a vision system does is computed from those numbers. The lecturer's conclusion is that computers are better and faster at easy things, such as comparing intensities and classifying images, while people are better at hard things, such as reading the context behind a picture.
People also have their own weaknesses, shown by a famous pair of faces and a moving illusion.
Computer vision against pattern recognition
The lecture closes with a comparison table. Computer vision extracts meaningful information from images and works only on visual data such as images and video.
Pattern recognition identifies patterns and regularities in any data for classification, including text, audio and signals.
Computer vision applications include driverless cars, medical diagnostics, crop and livestock monitoring and defect detection; pattern recognition applications include computer-aided diagnosis, bioinformatics, data compression and signal processing.
Applications taught
The lecture's examples include 3D maps, computational photography, depth estimation, iris and fingerprint biometrics, object recognition from a phone photo, a virtual line drawn on a sports broadcast, motion-sensing game controllers, mixed-reality headsets, a Mars rover, smart cars with several cameras, and a drone that found a lost hiker on Kowloon Peak in 2018.
What this chapter covers
- 01
How much human information is visual
- 02
Definitions and a short history
- 03
Pixels as intensity numbers
- 04
Easy tasks for machines, hard tasks for people
- 05
The comparison table
- 06
Applications from maps to rescue drones
Worked example · free
Advise a hotel on a vision system
- 2Explain what the system processes.
- 2Name a task it will do well and why.
- 2Name a task it will do badly and why.
Key terms
- Computer Vision
- The study of how machines extract meaningful information from images and video.
- Pixel Intensity
- The number stored for each point of an image, higher for brighter points.
- Biometrics
- Identifying a person from a physical feature such as an iris, a face or a fingerprint.
- Computational Photography
- Using algorithms to improve or construct images, such as simulating a wider camera aperture.
- Object Recognition
- Identifying what an object in an image is, for example by matching it against images online.
- Augmented Reality
- Placing computer-generated objects into a person's view of the real world.
Computer Vision and Pattern Recognition FAQ
What is the difference between computer vision and pattern recognition?
Computer vision works only on visual input, images and video, and aims to understand features in them. Pattern recognition is broader: it identifies regularities in any data, such as speech, text, signals or images, in order to classify it. Every computer vision system recognises patterns, but a voice recogniser recognises patterns without seeing anything.
Are computers better than people at seeing?
The lecture's answer is that it depends on the task. Computers are faster and often better at easy things, such as comparing intensity values and sorting images into classes. People are much better at hard things, such as understanding the context and meaning behind a picture, because to a computer an image is only a grid of numbers.
Why does an image look like numbers to a computer?
Because a digital image is stored as a grid of pixels, and each pixel is a number for its brightness. In an 8-bit greyscale image those numbers run from 0 for black to 255 for white. Colour images store several such numbers per pixel. All recognition, from edges to faces, is arithmetic on these values.
Exam move
Learn the four-row comparison table between computer vision and pattern recognition until you can reproduce it, then sort ten applications from the lecture into the two columns. Finally, practise explaining one application in the easy-task and hard-task language, because that framing suits almost any vision question.
Working through Computer Vision and Pattern Recognition in MM1041? Sia is AskSia’s AI Information Technology tutor — ask any MM1041 Computer Vision and Pattern Recognition question and get a clear, step-by-step explanation grounded in how MM1041 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
