Notes Ip11

Embed Size (px)

Citation preview

  • 8/3/2019 Notes Ip11

    1/13

    (11)(11) Representation and DescriptionRepresentation and Description

    - Low-level image processing* Image enhancement, restoration, transformation

    Image ImageEnhancement

    EnhancedImage

    Restored/Transformed

    ImageRestoration/Transformation

    - Mid-level image processing (image understanding)* Object representation, description

    Restored/Transformed

    ImageImageSegmentation

    SegmentedImage

    Representation/Description/Features

    Object

    Representation/Description

    11-1

  • 8/3/2019 Notes Ip11

    2/13

    - High-level image processing (recognition and interpretation)* Object recognition, interpretation of object relationships

    ObjectRecognition

    Objects

    Meaning/

    RelationshipsImageInterpretation

    Representation/Description/

    Features

    (a)(a) Chain codeChain code

    - Chain codes: represent a boundary by a connected sequence of straight-line segments of specified length and direction* Choose an appropriate grid to approximate objects

    11-2

  • 8/3/2019 Notes Ip11

    3/13

    0

    3

    2

    1

    4-direction: 8-direction:

    04

    1

    2

    3

    56

    7

    2

    0 0

    3

    3

    3

    3

    3

    3

    22

    12

    1

    1

    10

    01

    1

    07

    6

    6

    6

    5

    53

    3

    2

    1

    2

    20

    11-3

  • 8/3/2019 Notes Ip11

    4/13

    * Chain code (clockwise): 4-direction: 00333332322121110101, 8-direction: 07666553321202

    * Problem 1: different starting points result in different chain codes# Solution: normalization redefine the starting point such that the chain

    code forms a smallest number# E.g.: 6553320000 0000655332

    * Problem 2: object rotation results in different chain codes# Solution: difference code coding with the difference of directions

    (counter-clockwise)

    # E.g.: 0000655332 0006706076 0006706076 (normalization)

    (b)(b) Polygonal approximationPolygonal approximation

    - Polygonal approximation: approximate a boundary using a polygon

    - Minimum perimeter polygons* Choose an appropriate grid The boundary is enclosed by a set of concatenated cells

    * Allow the boundary to shrink as a rubber band

    11-4

  • 8/3/2019 Notes Ip11

    5/13

    - Merging techniques

    1. Merge points along a boundary until the least square error line fit ofthe points merged so far exceeds a threshold

    2. Record the the two end point of the line3. Repeat Steps 1 and 2 until all boundary points are processed

    11-5

  • 8/3/2019 Notes Ip11

    6/13

    - Splitting techniques

    Subdivide a boundary segment successively into two parts until aspecified criterion is satisfied

    1. Find two points on the boundary that are farthest away and draw a line

    Splitting line divides the boundary into two boundary segments2. For each boundary segment, find a point on the boundary that has a

    maximum perpendicular distance to its corresponding line

    11-6

  • 8/3/2019 Notes Ip11

    7/13

    3. Draw two lines joining the point and the two end points, respectively,of the corresponding splitting line Subdivide the boundary segment into two parts

    4. Repeat Steps 2~3 until the perpendicular distance is less than a

    threshold

    11-7

  • 8/3/2019 Notes Ip11

    8/13

    (c)(c) SignatureSignature

    - Signature: a 1D functional representation of a boundary* Plot the distance from the centroid to the boundary as a function of

    angles:signature = r(), = 0 ~ 2

    (d)(d) Boundary descriptorsBoundary descriptors

    - Descriptions of boundary

    * Perimeter: length of the chain code* Diameter: maxi,j(D(pi,pj))

    11-8

  • 8/3/2019 Notes Ip11

    9/13

    pi

    pj

    Ma= max

    i,j(D(p

    i,p

    j))

    Ma

    ma

    and forms a

    basic rectangle

    Ma

    ma

    * Major axis (Ma) and minor axis (ma)* Eccentricity: Ma/ma

    - MATLAB p = bwperim(bw, conn): find perimeter pixels* bw: binary image* conn: 4 or 8 (connectivity)* Perimeter pixels: 1-valued pixels that are connected to at least one 0-

    valued pixels* p

    : returned perimeter binary image- MATLAB S = diameter(L): find descriptions of a boundary

    * L: a labeled image

    11-9

  • 8/3/2019 Notes Ip11

    10/13

    * S: a structure with the following fields# S.Diameter, S.MajorAxis, S.MinorAxis, S.BasicRectangle

    (e)(e) Regional descriptorsRegional descriptors

    - Compactness: perimeter2/area

    - Topological: Euler number,E= C H (number of connected component number of holes)

    E= 1 3 = 2 E= 1 1 = 0 E= 1 2 = 1

    - MATLAB D = regionprops(L, properties)

    11-10

  • 8/3/2019 Notes Ip11

    11/13

    * L: labeled image* Properties: 'Area', 'BoundingBox', 'Centroid'* D: a structure with the fields specified when invoking regionprops( )

    * E.g., D.Area

    - Texture* Statisticcal approaches

    * Intensity mean: m=i=0

    L1

    zip zi ,

    * nth moment about the mean: n=

    i=0L1

    zimnpzi

    Moment Expression Description

    Mean m=i=0L1

    zipzi Measure of average intensity

    Standard deviation =zz=2

    Measure of average contrast

    Smoothness R = 11/(1+2

    ) Measure of smoothness of intensity

    Third moment 3=i=0L1

    zim3pzi Measure of skewness of the

    histogram; 0: symmetric, positive:skewed to the right; negative:

    11-11

  • 8/3/2019 Notes Ip11

    12/13

    skewed to the left

    Uniformity U=i=0

    L1

    p2zi Measure of uniformity; maximum: all

    graylevels are equal

    Entropy e=i=0

    L1

    pzi log2 zi Measure of randomness

    * Sprectral approach: based on Fourier spectrum* Structural approach: structure of the texture primitives

    11-12

  • 8/3/2019 Notes Ip11

    13/13

    (f)(f) High-level image processingHigh-level image processing

    - High-level image processing (computer vision)* Recognition and interpretation

    * Object recognition: identify objects based on object models* Interpretation: using artificial intelligence to infer the following

    # Properties of objects: identity, size, material, 2D/3D position, orientation# Relationships among objects: occlusion, relative position

    Further inference: planning of path, operations, control

    - Applications: robotics, industrial automatic inspection, autonomousnavigation, document image analysis, bio-information recognition

    11-13