Image Processing
Overview
What is image processing?
Breakdown of Image processing
Related Terminology
Image processing algorithms
Point Process
Area Process
Geometric Process
Frame Process
Miscellaneous
What is image processing?
The science of manipulating a picture which has
been already captured or generated.
Typically deals with but not limited to 2 dimensional
data.
Employs mathematic or computational data
structures and theories including matrix, vector,
array, convolution, Fourier transformation, sampling,
and etc .
Breakdown of Image processing
Image enhancement
Image restoration
Image analysis
Fictional manipulation (photomontage)
Image enhancement
improving an image for human observer by adjusting
brightness, contrast, hue and saturation.
Image restoration
restores image to some pre

specified quality
Image analysis
analyzing image for useful content such as counting
a number of blood cells in an image, finding and
tracking missile and finding defects in an product line
Fictional manipulation
(photomontage)
creation of an image
assembled from
different source
images for artistic,
commercial and
political purposes.
Artwork by Paul Allister
Related Terminology
Matrix
Array
Threshold
spatial/frequency domain
Convolution
Fourier Transformation
Interpolation
Matrix
An
M*N matrix
, denoted by
A
, is a rectangular array of
elements( usually numbers) enclosed typically by square
bracket.
M
is the number of horizontal rows and
N
the number of
vertical columns in the array.
A =
Matrix
Matrix has its own way of operation( adding, subtracting,
multiplication & division).
Many computational data structures are derived from, or related
to this notation. (ex. Database, raster (bitmap) image)
0 2
3 4
1 2

1 3
1 4
2 7
+ =
Threshold
Definition:
n 1. the starting point for a
new state or experience
2: the smallest detectable
sensation

Source:
WordNet ® 2.0, ©
2003 Princeton University
Used as a parameter value
for many digital imaging
processing.
Image as in a frequency/spatial
domain
While spatial domain looks at a picture as a collection of
brightness levels
frequency domain looks at a picture as a collection of
frequencies.
‘Spatial frequency’ of an image refers to the rate at which
the pixel intensities change.
Higher frequency image has more variations in the
brightness of pixels than lower frequency image
Image as in a frequency/spatial
domain
Image processing algorithms
Point process
: modifies a pixel’s value based on that
pixel’s original value or position
Area process
: Modifies a pixel’s value based on its
original value and the values of neighboring pixels.
Geometric process
: changes the position or
arrangement of the pixels
Frame process
: Generates pixel values based on the
operation on two or more images
Point Process
Operates on a pixel based solely on that pixel’s
value(simplest)
Adding, subtracting, dividing, multiplying pixels by a
constant value
Photoshop

Level, curve
Point Process

Arithmetic operation
Adding, subtracting, dividing, multiplying pixels by a constant
value
P1 = P0 + k ( brighter)
P1 = P0
–
k (darker)
P1 = P0 * k (higher contrast)
P1 = P0 / k (lower contrast)
Point Process

Arithmetic operation
+ 50

50 *2 /2
original
Point Process

Histogram

A bar graph of the pixel intensities.

A valuable tool used to view the intensity profile of an
image.
Histogram
Point Process

Histogram
Histogram Equalization

Redistribution of pixel intensity to achieve uniform distribution
Contrast Stretching

Image with good contrast exhibit a wide range of pixel
values.

Contrast stretching is applied to an image to stretch a
histogram to fill the fully dynamic range of image (adjusting
slides at level tool)
Intensity Transformation

a point process that converts an old pixel into a new pixel
based in some predefined function( curve tool)
Histogram Equalization
Image processing algorithms
Point process
Area process
Geometric process
Frame process
Area Process

Use the input pixel as well as the pixels around it to
generate a new output.

Used in various filters such as blur, sharpen, emboss,
edge detection, median filter, and etc…
Area Process

Convolution
A weighted sum of pixels in the neighborhood of the
source pixel
The weights are determined by a small matrix called
the convolution mask or convolution kernel.
The location of the center corresponds to the location
of the output pixel
The sum of weights in the convolution mask affect
the overall intensity of the resulting image.
Various convolution masks for each filter can be
experimented at Photoshop
Area Process

Convolution
Area Process

Noise correction

Noise can be generated while digitization

Noise can be removed by using noise filter technique(Low pass
filter and median filter)

Median Filter good for correcting impulsive nose, Low pass filter
is good for removing Gaussian nose in the electronic
applications.
Area Process

Median Filter
Area Process

Fourier
Transformation
In order to represent data in the frequency domain,
Fourier transformation is necessary.
Jean Baptiste Joseph (1807) has proposed the idea
that any periodic signal could be represented by a
series of sinusoids. (ex. Prism, sound, and even
bitmap image)
A transformation of a signal into a series of different
ranges of frequencies.
Types: DFT, FFT…
Area Process

Fourier
Transformation
Image Filtering in the
frequency domain
FFT

> Manipulation

>Inverse FFT
Image processing algorithms
Point process
Area process
Geometric process
Frame process
Geometric Process

Modifies the arrangement of pixels based on some
geometric transformation

Scale, rotation, move, mirror

Employs interpolation method to create a new pixel
based on the source pixel and the transformation
functions. (ex. Resampling at Photoshop)
Image processing algorithms
Point process
Area process
Geometric process
Frame process
Frame Process

generates a pixel value based on an operation
involving two or more different images

The pixel operations generate an output image based
on the operation of a pixel from two separate image.
(Ex. blending mode at Photoshop
http://www.pegtop.net/delphi/blendmodes/
)
Miscellaneous
Alpha channel
A channel which is reserved for transparency information of an
image.
The alpha channel is a mask

it specifies how the pixel's colors
should be merged with another pixel when the two are overlaid,
one on top of the other.
An image can be saved using four channels

three 8

bit
channels for red, green, and blue (
RGB
) and one 8

bit alpha
channel.
Not all the image format supports alpha channel. (Tiff, Tga
>
JPG. GIF, PNG)
Not all the programs can recognize alpha channel.( AfterEffect,
Final Cut Pro
> Preview)
Alpha channel is useful when you want to create an effect
based on the transparency information of an image.
Miscellaneous
Fading effects at Photoshop
The Fade command changes the opacity and blending mode of
any filter, painting tool, erasing tool, or color adjustment.
Anti

aliasing
The blending of colors around the edges of an object in a
digital image to create smooth transitions between shapes
and background colors.
Aliasing
The hard and blocky looking edge on a computer generated
curve.
Image processing for the moving
image
You can apply image processing to each frame of the move
as the same way you do to a still image
The main differences between still and moving images is
added element of time manipulation or manipulation over
time.
References
A Simplified Approach to Image Processing by Randy Crane,
Prentice Hall 1997
http://www.ph.tn.tudelft.nl/Courses/FIP/noframes/fip.html
http://www.imageprocessingbook.com/DIP2E/image_databases/
image_databases.htm
http://web.uct.ac.za/depts/physics/laser/hanbury/intro_ip.html
http://www.pegtop.net/delphi/blendmodes/
http://www.thinkdan.com/tutorials/photoshop/blendingmodes/ind
ex.html
Lecture Note by Trinity A. Greer, RIT 2003
Comments 0
Log in to post a comment