Lecture 1 - Intro

One picture is worth more than ten thousand words

1. Core Definitions

Computer Imaging: The acquisition and processing of visual information by a computer. It is divided into two categories based on the ultimate receiver of the information: Computer Vision and Image Processing .
Digital Image: A representation of a two-dimensional image as a finite set of digital values.
Pixels: The individual picture elements that make up a digital image. Pixel values represent intensity or gray levels.
Digitization: Implies that a digital image is merely an approximation of a real scene.
Common Formats: Images can be grayscale (1 sample per point) or RGB (3 samples per point: Red, Green, and Blue).

The field transitions from simple image manipulation to deep understanding, broken into three levels:

Low-Level Process: Takes an image as input and outputs an image (e.g., noise removal, image sharpening).
Mid-Level Process: Takes an image as input and outputs attributes (e.g., object recognition, segmentation). Note: This course will stop at this level.
High-Level Process: Takes attributes as input and outputs understanding (e.g., scene understanding, autonomous navigation).

Image Processing: Focuses on processing images for human consumption (Image In $\to$ Image Out).
Computer Vision: Focuses on processing output images for computer use .
Image Analysis: Extracts measurements from an image to solve vision problems (Image In $\to$ Measurements Out). It involves feature extraction (acquiring high-level info) and pattern classification (identifying objects) .
Image Understanding: Maps an image to a high-level description (Image In $\to$ High-level description Out).

1920s: Early applications were in the newspaper industry. The Bartlane system used submarine cables to transfer 5-level coded pictures between London and New York . This was later improved to 15-tone images.
1960s: The space race and improvements in computing led to a surge in DIP. Computers were used to improve images of the moon taken by the Ranger 7 probe in 1964.
1970s: DIP expanded into medical applications. Notably, Sir Godfrey N. Hounsfield and Prof. Allan M. Cormack won a Nobel Prize in 1979 for inventing tomography, leading to CAT scans.
1980s - Today: The use of DIP techniques has exploded across various industries.

Digital image processing is utilized in highly diverse fields :

Image Enhancement: Improving quality and removing noise, famously used to correct the early images from the Hubble Telescope .
Medicine: Analyzing MRI scans to find boundaries between different tissue densities .
Artistic Effects: Used to make images more visually appealing ,to add special effects and to make composite image.
GIS (Geographic Information Systems): Manipulating satellite imagery for terrain classification, meteorology, and mapping night-time lights of the world .
Industrial Inspection: Using machine vision to replace slow, expensive human operators, such as checking printed circuit boards (PCBs) for missing components and acceptable solder joints .
Law Enforcement: Utilizing automated number plate recognition, fingerprint recognition, and CCTV enhancement .
Human-Computer Interfaces (HCI): Developing natural interfaces via face and gesture recognition .

A complete DIP pipeline involves several interconnected stages:

Image Acquisition: The initial capture of the visual problem domain.
Image Enhancement: Taking an image and improving its visual appearance.
Image Restoration: Reversing known or estimated degradation to return an image to its original appearance .
Morphological Processing: Extracting image components useful for representing shape, such as boundaries, skeletons, and convex hulls .
Segmentation: Subdividing an image into its constituent parts or objects.
Object Recognition: Identifying the segmented objects within the image.
Representation & Description: Describing the recognized objects for further processing.
Image Compression: Reducing the massive amount of data required to store or transmit an image.
Color Image Processing: Handling the unique properties of colored components.