Notes 20110113 CIS 6320 Image Processing

From SnOwy - Ed's Wiki Notebook

• notes are given in packets
• digital image processing: introduction and fundamentals

Contents 1 Origins 2 Electromagnetic Spectrum -- and Imaging 3 Applications 4 Recall 5 Physiological 6 Lancet(sp) 7 (seven channels) -- satellite image 7 Imaging Modalities -- non electro magnetic 8 Image Processing Overview 9 Partitions

Origins

• 1920s -- from the newspaper industry
• 1921 newspaper pictures were sent by submarine cable (London to New York) -- only 5 to 15 grey levels
• 1964 -- picture of the moon from US spacecraft

Electromagnetic Spectrum -- and Imaging

• λν = c -- wavelength * frequency
• E = hν -- Plunk's relation -- h is Plunk's constant

Applications

• nuclear medicine -- imaging with γ-rays
• circuit board -- x-rays
• fluorescent microscope images of corn -- ultraviolet
• paper currency serial numbers -- visible spectrum
• satellite image, human satellite -- infrared
• MRI sport-related injuries -- radio-band
• satellite image, cities (1:25000; 1:10000; "resolution is 5m") -- panchromatic: greyscale given a collapsure of 480 to 710nm
• spatial resolution is given by number of meters per pixel (single-dimension)

Recall

• short wavelength is blue
• long wavelength is red
• wavelength is inversely proportional to frequency -- as long as c stays common

Physiological

• we must be aware of how humans process images
• compression and image reproduction -- overlaps with psychology and cognitive science

Lancet(sp) 7 (seven channels) -- satellite image

• 450-515nm blue: water
• 525-605nm green: vegetation
• 630-690nm red: soil
• 775-900nm near-infrared: soil?
• 1550-1750nm middle-infrared: moisture
• 10400-12500nm far-infrared: heat
• 2090-2350nm middle-infrared: geological
• multispectral image
• false colour image -- assignment of channel to a visible colour (or a different visible colour).

Imaging Modalities -- non electro magnetic

• Ultrasound
• Synthetic image

Image Processing Overview

• IP is classified depending on the goal of the processing
• low level -- image preprocessing
  1. imaging acquisition -- the very first step of the processing
  2. reprocessing
     • image restoration -- based on objective grounds -- removing artefacts given known physical processes affecting the image
     • image enhancement -- based on subjective grounds -- we are modifying an image aside from the information content
  • the outputs from these low level steps are images
• knowledge base
  • if we know something about an artefact given where or how the image was captured, the knowledge base informs image restoration
• high level
  1. segmentation -- partition the image into meaningful segments
  2. representation and description
  3. object recognition

Partitions

• U is the universe of discourse (the image)
• Let S be the partition of U ...
• Let U_i be a subset of U ...

1. ∀i in |S|, ∀j in |S|, i ≠ j → U_i intersect U_j = Φ
2. Union _i=1^|S| U_i = U
3. ∀i in |S|, U_i ≠ Φ

• note: "S" on the whiteboard has an over curve concave down -- like a person wearing half a watermelon for a hat.