david e. pitts csci 5532 overview of image processing by david e. pitts aug 22, 2010 copyright 2005,...

25
David E. Pitts CSCI 5532 Overview of Image Processing by David E. Pitts Aug 22, 2010 copyright 2005, 2006, 2007, 2008, 2009, 2010

Upload: poppy-booker

Post on 02-Jan-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

David E. Pitts

CSCI 5532CSCI 5532

Overview of Image Processing

by David E. Pitts

Aug 22, 2010copyright 2005, 2006, 2007, 2008, 2009, 2010

David E. Pitts

CSCI 5532CSCI 5532

A Digital Image is an Array of Numbers Spatially Arranged in a Predetermined FashionEach number in the Array represents one Picture Element

One Pixel

David E. Pitts

CSCI 5532CSCI 5532

Digital Image

-Light intensity across the image is a function of x and

-grey level range in each pixel is often 8 bits (256 grey levels)(the human eye resolves about 32)

David E. Pitts

CSCI 5532CSCI 5532

Sources of Images

- Film Cameras (film is sensitive from 0.3 to 0.9 µm)-transparencies-prints

-Video Cameras (silicon is sensitive into near infrared 1.1 µm)-X-rays (e.g. Computerized Tomography)-MRI (Magnetic Resonance Imaging)-Ultrasound (medical images)-Infrasound-Seismic-Aircraft and Spacecraft scanners-Telescopes

-Radio, Optical, Gamma Ray, X-ray-Positron Emission Tomography

David E. Pitts

CSCI 5532CSCI 5532

Image Artifacts

-Random noise (e.g. blank TV screen)

-Periodic Noise (e.g. herringbone patterns)

-Poor gain (image is too bright or too dark)

-Poor contrast (image is too uniform in brightness)

-Nonuniform response across image (e.g. corners darker than center)

-Smear (e.g. subject moving during image acquisition)

-Curvature of field (e.g. fish eye lens)

- Poor Color

David E. Pitts

CSCI 5532CSCI 5532

Approach for Removing Artifacts

- Use off the shelf applicaitons- Photoshop http://www.adobe.com/

- NIH Image http://rsbweb.nih.gov/nih-image/

- Multispec http://dynamo.ecn.purdue.edu/~biehl/MultiSpec/

- Dimple http://www.process.com.au/AboutDIMPLE.shtml

- Develop your own algorithms based on fundamental principles

David E. Pitts

CSCI 5532CSCI 5532

Mathematical Operators Used in Imaging

AdditionSubtractionMultiplicationDivision

Linear TransformsNon-Linear Transforms

DifferentiationIntegration

Fourier Transforms

Morphological Operators

Boolean Operators

David E. Pitts

CSCI 5532CSCI 5532

Because of the Nature of Digital ImagesOperators are Discrete (e.g. Integration is replaced by Summation)

Operations can be:Between Two or More ImagesOver an Entire ImageWithin a Window ≥ 2 x 2 pixels

David E. Pitts

CSCI 5532CSCI 5532

Operations Can Involve Any One of Many Characteristics of an Image

Grey Scale Content (Brightness and Contrast)Color Content (Hue and Saturation)Geometry of Image (2 D, 3 D)Frequency Content of Image (e.g. Noise, Edges, Texture)Shapes of FeaturesAssociation of Patterns (e.g. Shadows)

David E. Pitts

CSCI 5532CSCI 5532

Purpose of Imaging Processing

Correct Image DefectsBlurringDistortionsImprove Brightness or ContrastImprove ColorMinimize NoiseRemove Non-Uniform Lumination

David E. Pitts

CSCI 5532CSCI 5532

Purpose of Imaging Processing (cont.)

Segmentation of Image (Isolate Desired Features in order to determine characteristics)

SizeLocationShapeOrientationPerimeterMajor and Minor Dimensions

David E. Pitts

CSCI 5532CSCI 5532

Single Pixel Operations Within One Digital Image

Operation

David E. Pitts

CSCI 5532CSCI 5532

Single Pixel Operations Within One Digital Color Image

Operation

RG

B

RG

B

David E. Pitts

CSCI 5532CSCI 5532

Linear Algebraic Operations Between Digital Images

Operation

X Y Z

David E. Pitts

CSCI 5532CSCI 5532

Operations Within Windows in One Digital Image

Operation

David E. Pitts

CSCI 5532CSCI 5532

Operations Within Windows in One Digital Image

Example: Edge Enhancement of the function g(x,y)

Laplacian =

David E. Pitts

CSCI 5532CSCI 5532

Operations Within Windows in One Digital Image

Median Filter = Sort Values from Pixels in Window

David E. Pitts

CSCI 5532CSCI 5532

Operations Within Windows in One Digital Image

Average Value Filter

David E. Pitts

CSCI 5532CSCI 5532

TEXTURE INFORMATION TECHNIQUES

GRAY LEVEL STATISTICS

FOURIER POWER SPECTRUM

GRAY LEVEL DIFFERENCE STATISTICS

DECORRELATION METHODS

MARKOV RANDOM FILED

David E. Pitts

CSCI 5532CSCI 5532

3 X 3 WINDOW WITH GRAY LEVEL 0 TO 3

1 2 3

1 1 2

0 1 2

David E. Pitts

CSCI 5532CSCI 5532

A GREY LEVEL CO-OCCURRENCE MATRIX (P )IS DEVELOPED FOR THE DESIRED ORIENTATIONHORIZONTAL, VERTICAL, OR DIAGONAL.

i j

REPRESENTS THE NUMBER OF TIMES i and jARE NEIGHBORS

0 1 2 3 i

0

1

2

3

j

David E. Pitts

CSCI 5532CSCI 5532

MANY MEASURES OF TEXTURE CAN BE USED

ANGULAR SECOND MOMENT

∑ { P /R }

ENTROPY

∑ { P /R } LOG { P /R }

2

i j

i j

i j

i j

i j

David E. Pitts

CSCI 5532CSCI 5532

INVERSE DIFFERENCE MOMENT

∑ { P /R }/{1/(1+(I-J) )}

AD HOC CHOICES OF # OF GREY LEVELS,ANGULAR ORIENTATION AND SIZE OF WINDOWMUST BE MADE ACCORDING THE TEXTURE TOBE DETECTED

i ji j

David E. Pitts

CSCI 5532CSCI 5532

Multiple viewing angle videography

Developed by Carnegie Mellon School of Computer Science

Technique was first used at 2001 Superbowl

Technique uses many synchronized video cameras arrangedin a circle around the scene. At any given time morphing techniques can be utilized to “fly around the scene”.

http://www.cs.cmu.edu/virtualized-reality/main.html

David E. Pitts

CSCI 5532CSCI 5532

Image Processing Software

Photoshop by Adobe is the most widely used software. Cost $599Photoshop Elements by Adobe contains most commonly used features of

Photoshop. Mac & Windows, Cost $89

Other Image Processing Software was reviewed in the March/April 2003 issue of Photoelectronic Imaging, pg 14-15.

Asiva Photo by Shapiro Consulting - Windows $378LuminaXYs by Tetrix Technologies - Mac & Windows $99Dfine Photoshop plug-in for noise reduction www.nikmultimedia.comiCorrect Professional 4.0 by Pictographics Mac & Windows $79PhotoKit by Pixel Genius is a Photoshop plug-in, Mac & Windows $49