# Image Processing

Τεχνίτη Νοημοσύνη και Ρομποτική

6 Νοε 2013 (πριν από 4 χρόνια και 6 μήνες)

123 εμφανίσεις

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

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:
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

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