Scribd
Upload
15 views18 pages

4 - GGI 3203 - Introduction To Object Based Image Analysis - 2023

Uploaded by

kiptoo.bethwel21
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
15 views18 pages

4 - GGI 3203 - Introduction To Object Based Image Analysis - 2023

Uploaded by

kiptoo.bethwel21
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 18

GGI 3203

Object Based Image Analysis (Segmentation)


Content:

• Pixel-based approaches
• Benefits
• Challenges
• Segmentation
• Methods
Pixels vs Objects: Pixel based
- Pixel based classification methods are also referred to
as spectral or point classifiers
- Point classifiers because they consider each point/ pixel
as an observation (i.e. as values isolated from their
neighbors).
- Spectral classifiers because comparison of pixels is
done in the feature space (scatter plot/ scatter gram)
- If pixels are compared to one another in the feature space,
we can group similar pixels together
- Pixels in one class are spectrally different from other pixels in
another class.
Pixels vs Objects: Pixel based
- Benefits of pixel based approach:
- Simplicity
- Economy

- Problems:
- Spectral values belong to more than one information class
- No spatial relationships used in classification (Pixel is an
artificial spatial unit)
- ‘Artifacts’ (salt-and-pepper effect)
Pixels vs Objects: Object scope
Manual
delineation

Pixel based Object based


classification classification
Pixels vs Objects: Object scope
- The basic elements of an object-oriented approach are
image objects.
• Image objects are:
• Groups of pixels that are similar to one another.
• Contiguous regions in an image.

- Image objects are found by image segmentation using


• spectral properties,
• size,
• shape,
• texture, and
• maybe other context from a pixel neighborhood.
Image Segmentation
Image segmentation
• Division of an image into regions so that
• the whole scene (exhaustive segmentation) is
covered by regions (spatially continuous,)
• the regions do not overlap
• the regions are homogeneous within themselves
• the homogeneity criteria of neighboring regions
differ
Object based image analysis
Image segmentation: Methods include
• Pixel-based or histogram-based
• Thresholding techniques
• Segmentation in the feature space
• Region-based
• Region growing, merging and splitting, watershed,
• Edge-based
• Laplace filter, Sobel-operator etc
Image segmentation: Methods: Histogram Thresholding
• Histogram Thresholding
• Simplest way to accomplish exhaustive regional
segmentation
• The spectral feature space is separated into subdivisions, and
pixels of the same subdivision are merged when locally
adjacent in the image data.

• Limited Quality
• Oversegmentation
• Undersegmentation
Image segmentation: Methods: Region growing
• Region Growing
• Seed cells are distributed over image
• Neighbors (4- or 8-neighbourhood)
are included into region, if
• they do not belong to another
region yet
• the homogeneity criterion H
applies
• Two neighboring regions are unified,
if H applies
Image segmentation: Region growing by pixel aggregation

0 0 5 6 7
1 1 5 8 7
0 1 6 7 7
2 0 7 6 6
0 1 5 6 12

Homogeneity criteria: <3


Homogeneity criteria: <8?
Image segmentation: Methods: Region growing
• Region Growing
ECHO [extraction and classification of homogeneous objects]
- searches for neighbouring pixels that are spectrally similar, then
enlarges these groups to include adjacent pixels that have spectral
values that resemble those of the core group.
- For example, the algorithm can first search for neighborhoods of four
contiguous pixels. For each group, it then tests members for
homogeneity […] Pixels that are not similar to their neighbours are
rejected from the group […].
- Each of the homogeneous patches is then compared to each of its
neighbors. If similar patches border each other, they are merged to
form a larger patch.
- Patches are allowed to grow until they meet the edges of contrasting
patches; when all patches reach their maximum extent within the
constraints defined by the operator, the growing process stops
Image segmentation: Methods: Region based
• Split and Merge’
• Combination of coarse segmentation and merge
• Example: Quadtree
• Initially: image is considered as one object
• division into 4 parts, if H does not apply
• Resulting quadtree structure
• Merge of homogenous quadtree areas
Image segmentation: Methods: Region growing
Image segmentation: Methods: Edge based segmentation

• Edge-based segmentation
• Edge: boundary between homogenous areas
• Region-based segmentation makes sense when large, compact and
coherent objects occur
• Edge-based segmentation for elongated structures
Image segmentation: Methods: Edge based segmentation
• Workflow for Edge based segmentation
1.Edge detection
• Filtering – smoothing to decrease noise in the image
• Enhancement – revealing local changes in intensities
• Detection – select edge pixels, e.g. by thresholding
• Closing of gaps / deleting artefacts
• Combining, extending of lines
2.Linking the edge pixels to form the region boundaries

• Enhancement filters
• Sobel operator
• Laplace filter
• Compass edge

You might also like