Top Banner

Click here to load reader

Image Processing - Ch 10 - image segmentation - ECE/CIS barner/courses/eleg675/Image Processing... · PDF fileImage Segmentation Image Processing with Biomedical ... Image Processing

Mar 22, 2018

ReportDownload

Documents

nguyenlien

  • 1

    Image Segmentation

    Image Processing with Biomedical Applications

    ELEG-475/675Prof. Barner

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 2

    Image Segmentation

    Objective: extract attributes (objects) of interest from an image

    Points, lines, regions, etc.Common properties considered in segmentation:

    Discontinuities and similaritiesApproaches considered:

    Point and line detectionEdge linking

    Thresholding methodsHistogram, adaptive, etc.

    Region growing and splitting

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 3

    Detection of Discontinuities

    Mask filtering approach:

    Isolated point detection: |R|TExample:

    X-ray imageT=90%of maxvalueInput,gradient,thresholdoutput

    9

    1 1 2 2 9 91

    i ii

    R z z z z =

    = + + + =

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 4

    Line Detection

    Line detection masksDetects lines one pixel wideLine orientation specific

    Set orientations specific thresholdsSecond derivative based

  • 2

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 5

    Line Detection Example

    Wire-bond mask for electronic circuitApplication of -45 edge maskResult of thresholding

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 6

    Edge Detection

    Concepts:Edge localBoundary global

    Ideal edge:Step

    Practical edge:RampIdeal edges are smoothed by optics, sampling, illumination conditionsInch thickness determined by transition region

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 7

    Edge Example Noiseless Case

    Ramp edgeThe first derivative:

    PulseThick edges

    Second derivative:Spikes at onset and terminationZero crossing marks edge center

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 8

    Edge Example Noisy Case

    Gaussian noise corrupted edgeDerivatives amplify noise

    Even modest levels of noise severely degraded gradient-based edge detectionPossible solution: noise smoothing prior to edge detection

  • 3

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 9

    Gradient Operators

    Two-dimensional gradient:

    Magnitude:

    Direction (angle)

    Perpendicular to edgeApproximation:

    Shown: mask realizations

    x

    y

    fG x

    fGy

    = =

    f

    ( )1/ 22 2

    x yf mag G G = = + f

    1( , ) tan yx

    Gx y

    G

    =

    x yf G G +

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 10

    Gradient Operators and Example

    Application: Horizontal, vertical and (additive) gradient

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 11

    Gradient Example (I)

    Preprocess image

    Smooth detail texturesThicken edgesFilter: 5x5 averaging filter

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 12

    Gradient Example (II)

    Extension to 45gradients and their application

  • 4

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 13

    Laplacian of a Gaussian (LoG)

    Recall Laplacian:

    Edge detection limitations:Produces double edges, insensitive to edge direction, sensitive to noise

    Pre-smooth withGaussian filter

    Combined (linear) smoothing and derivative operations

    2 22

    2 2

    f ffx y

    = +

    2

    22( )r

    h r e

    =

    2

    22 2

    2 24( )

    rrh r e

    =

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 14

    LoG Example

    Angiogram exampleSobel output shown for reference

    To obtain edges:Threshold LoGMark zero crossings

    Numerous false (spaghetti) edgesFirst derivative more widely used

    LoG models certain aspects of the human visual system

    Image ProcessingImage Segmentation

    Prof. Barner, ECE Department, University of Delaware 15

    Edge Linking

    Procedures often yield broken edgesNoise, illumination irregularities, etc.

    Link neighboring segments based on predefined criteria

    Example criteria:Strength of gradients

    Direction of gradients

    Applied over predefined search neighborhood

    0 0( , ) ( , )f x y f x y E

    0 0( , ) ( , )x y x y A n (1) < n (0)

    [ ] { }( , ) | ( , )T n s t g s t n=

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.