Image Analysis - Pattern Recognition

The study of pattern recognition deals with techniques such as automatic recognition or the identification of images or elements of images. Pattern recognition means that a computer is able to respond to stimuli from an external world through the medium of sensors.

Humans are great at pattern recognition without even trying:

• Recognising faces.
• Understanding written text.
• Navigating through physical spaces.
• Hand eye coordination.
• Identification of food.

Pattern recognition aims to provide computers with this intelligence.

Image pattern recognition also provides many applications:

• Medical Image analysis
  • (Early signs of cancers)
  • Xray anomalies
• Text Analysis
• Self-driving cars
• etc.

Definitions:

• Pattern

  • Description of some object or entity in terms of the existence of identifying characteristics.

  • A pattern is represented as a feature vector where each element x represents a Feature or pattern descriptor.

  • We use data and make measurements to define patterns.

• Pattern Recognition

  • Overall process which allows the naming of a general category of object in response to data which form a specific pattern

• Pattern Class

  • Category determined by some common attributes among its members
  • The same object or entity may be categorised in different ways. Often depending on the application of interest.
  • Classes may be labelled in different ways:
    • Specifically: C1=‘A’, C2=‘B’
    • Broadly defined: C1 = Normal, C2 = Abnormal

• Feature

  • (Pattern descriptor): a measurable property of a pattern selected to contribute to its identification

Pattern Recognition Task