2
Comparison of different Segmentation Techniques S.No Segmentation Technique Information used in Segmentation Extends to N-D Method of Segmentation Advantages Disadvantages 1 Thresholding Colour No Simple Comparison Between Pixels And The Threshold When the intensity of the pixels is larger/smaller than a predefined threshold, those pixels are classified as foreground. Otherwise, they will be viewed as background. Threshold based approaches are the simplest, easiest and fast ones among all of the existed segmentation methods. The difficulty is that it is not easy to find an appropriate threshold which can separate the image into two groups directly. This method also requires the foreground and background in the image have obviously different intensity values. 2 Live Wire Edge No 2D Graph Search Using Dynamic Programming More general, reduced initialization. Real-time: User immediately sees whether desired contour was obtained or not and can adjust contour otherwise. Direct control of contour. Sub-pixel accuracy not possible. Contour depends on weighting between internal and external costs. May ignore jagged edges or snap to wrong edges. 3 Snakes/Activ e contours Edge No Energy Minimization of Internal and External Functions They autonomously and adaptively search for the minimum state. External image forces act upon the snake in an intuitive manner. Incorporating Gaussian smoothing in the image energy function introduces scale sensitivity. They can be used to track dynamic objects. They are sensitive to local minima states, which can be counteracted by simulated annealing techniques. Minute features are often ignored during energy minimization over the entire contour. Their accuracy depends on the convergence policy. 4 Graph cut Colour and Edge Yes Max- Flow/Min- Cut Algorithm For Energy Minimization The energy function is constructed based on regional and boundary information and it can achieve globally optimal result. Taking the sorce and sink points and grouping the points into source region and sink region automatically with out human interaction is difficult.

comparison of different methods.pdf

Embed Size (px)

Citation preview

  • Comparison of different Segmentation Techniques

    S.No Segmentation

    Technique

    Information

    used in

    Segmentation

    Extends

    to N-D

    Method of

    Segmentation

    Advantages Disadvantages

    1 Thresholding Colour No Simple

    Comparison

    Between

    Pixels And

    The

    Threshold

    When the intensity of the

    pixels is larger/smaller than a

    predefined threshold, those

    pixels are classified as

    foreground. Otherwise, they

    will be viewed as

    background.

    Threshold based approaches

    are the simplest, easiest and

    fast ones among all of the

    existed segmentation

    methods.

    The difficulty is that it is

    not easy to find an

    appropriate threshold

    which can separate the

    image into two groups

    directly.

    This method also requires

    the foreground and

    background in the image

    have obviously different

    intensity values.

    2 Live Wire Edge No 2D Graph

    Search Using

    Dynamic

    Programming

    More general, reduced

    initialization.

    Real-time: User immediately

    sees whether desired contour

    was obtained or not and can

    adjust contour otherwise.

    Direct control of contour.

    Sub-pixel accuracy not

    possible.

    Contour depends on

    weighting between internal

    and external costs.

    May ignore jagged edges

    or snap to wrong edges.

    3 Snakes/Activ

    e contours

    Edge No Energy

    Minimization

    of Internal

    and External

    Functions

    They autonomously and

    adaptively search for the

    minimum state.

    External image forces act

    upon the snake in an intuitive

    manner. Incorporating

    Gaussian smoothing in the

    image energy function

    introduces scale sensitivity.

    They can be used to track

    dynamic objects.

    They are sensitive to local

    minima states, which can

    be counteracted by

    simulated annealing

    techniques. Minute

    features are often ignored

    during energy

    minimization over the

    entire contour.

    Their accuracy depends on

    the convergence policy.

    4 Graph cut Colour and

    Edge

    Yes Max-

    Flow/Min-

    Cut

    Algorithm

    For Energy

    Minimization

    The energy function is

    constructed based on regional

    and boundary information and

    it can achieve globally

    optimal result.

    Taking the sorce and sink

    points and grouping the

    points into source region

    and sink region

    automatically with out

    human interaction is

    difficult.