UNIT 2: Image Processing Basics Flashcards
Analog to digital conversion
Step 1: Scanning
Step 2: Sampling
Step 3: Quantization
Step 4: Coding
Step 1: Scanning
• Field of View is divided into a matrix
• Each pixel is given a coordinate location or address
Step 2: Sampling
• Samples are taken of the data
• Sampling Frequency: number of samples taken per unit length
• Determined by the Pixel Pitch
Step 2 sampling: Pixel pitch
The distance between the center of two adjacent pixels
Step 2 sampling: Nyquist Theorum
The sampling frequency must be at least twice the highest frequency of the sine wave
• Ex: A wave with a frequency of 5 Hz MUST have at least 10 samples taken by the ADC.
Step 2 sampling: Moire Pattern
-Sampling error
-If the sampling rates are too low, specific artifacts occur called moire patterns
T or F: the more samples the better
True
Step 3: Quantization
a gray shade is assigned to each individual image pixel
Step 3 Quantization: Bit depth
-Range of gray shades available for the computer to “choose from”
-All the possible shades
-Represented in powers of 2:
2^4= has 16 shades
2^5= has 64 shades
2^16= has 4096 shades
Step 3 Quantization: Dynamic Range
All the chosen shades of gray
Step 3 Quantization: Grayscale
All the visible shades of gray
Step 4: Coding
The computer assigns binary code that represents that grayshade in that pixel by its location on the matrix
What is the purpose of digital image processing?
-Prepare: Correct for inherent inconsistencies and shift grayscale into human vision
-Optimize: Enhance/suppress anatomic details
-Analyze: Computer Aided Diagnosis (CAD)
Pre-Processing
corrections made to the raw digital image that are designed to normalize the image, preparing
Post-processing
refinements to digital image that are targeted towards the specific anatomy, to make it look how we want it to
Pre-processing: acquisition processing
corrections made to raw image data to correct for
What does pre-processing correct for?
- Inconsistencies in the x-ray beam
- Electronic noise
- Non-uniformity in screen thickness (CR)
- Non-uniformity in Laser scanning (CR)
- Accumulated Background Exposure
- Faulty Pixels or Detector elements (DEL)
Post-Processing
• All manipulation of the digital image data made after corrections have been made for data acquisition
• Default processing
• Operator functions
Domain Processing
How the computer sorts the data for efficient processing
What are the 3 processing domains
-Spatial domain
-Intensity domain
-Frequency domain
Spatial Domain
Sorts the information by location on the Detector
•Sorts image data by location of pixels in space
•Radiographs start and end in the spatial domain
•Spatial Processing tasks:
• Point processing
• Area/Local processing
• Global processing
Intensity Domain
Sorts the information by pixel value (gray shade)
• Histogram construction and analysis
• Application of the LUT
Frequency Domain
Sorts the information by size of structure, according to the frequency of signal
• Small structures generate high frequency
• Large structures generate low frequency
*high pass filter
*low pass filter
3 types of processing algorithms that occur in the spatial domain
-Point processing
-Area/Local processing
-Global processing
Point processing
Perform tasks on individual pixels
Area/Local processing
-Kernel: A matrix within a matrix
-Convolution Kernels: Applies the same weighted calculation to each pixel in the kernel
-Interpolation Kernels: Takes the average of the pixels and applies that value to the center pixel
Global processing
Massive spatial function across entire image
What is a Kernel? What are 2 common types of kernels?
-A matrix within a matrix
-Interpolation kernel and convolution kernels
Convolution Kernels
Applies the same weighted calculation to each pixel in the kernel
Interpolation Kernels
Takes the average of the pixels and applies that value to the center pixel
Histogram
• A graphical representation of the pixel values in an image (gray shades within an image)
• Histogram selected by radiographer
In what processing domain is the histogram constructed and analyzed?
Intensity domain
Fourier Transform
Algorithm that transforms image data from spatial domain to frequency domain
What processing domain is the Fourier Transform used in?
Frequency domain
Reverse Fourier Transform
Restacks the individual frequencies within the image data into a single frequency pattern
2 types of Frequency domain processing algorithms
-High pass filter
-Low pass filter
High pass filter
High frequencies “pass” onto the image and low frequencies are suppressed
• Edge Enhancement Filters
Low pass filter
Low frequencies “pass” onto the image and high frequencies are suppressed
• Smoothing Filters
Processing Algorithms
*computer may perform these tasks in multiple domains
-Image to Image transformations: Grayscale shifting
-Image to information transformations: Fourier transform
-Information to image transformations: used to display processed data
The smaller the pixel pitch the _________ the resolution
greater
Sample
representative part of a population, used for determining the characteristics of the whole population
sampling frequency or pitch
The number of samples taken per unit of length
sampling frequency is measured in
pixels/mm
Sampling Pitch
space between samples
