Image Processing

organizationduleAI and Robotics

Nov 6, 2013 (3 years and 1 month ago)

88 views

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