Image segmentation in erdas imagine

Pkm002-80-4.0
Image Segmentation in
ERDAS IMAGINE

2
Image Segmentation
▪ Segmentation is a way of
partitioning raster images
into segments based on
pixel values and locations.
Pixels that are spatially
connected and have similar
values are grouped in a
single segment.
Source: Clarklabs.org

3
Image Segmentation in ERDAS IMAGINE
Image Segmentation FLS Image Segmentation
Raster > Unsupervised

4
Image Segmentation in ERDAS
▪ Image Segmentation performs
edge detection on the raster image
first. Then it performs
segmentations on that raster image
using edges found in the edge
detection phase as boundaries of
segments. The result is a thematic
image in which pixel values
represent class IDs of contiguous
raster objects.
http://www.acgeospatial.co.uk/

5
Image Segmentation
• Use: Specifies how pixels
values are used in performing
segmentation.
• Intensity: Averages of all
band values are used in
performing segmentation.
•
All Layers: All results are
intersected to generate final
results; when this option is
selected, the Euclidean
Distance option becomes
available.

6
Image Segmentation Edge Detection

7
Image Segmentation
Edge Detection
▪ Pre Smoothing: Specify the
number of times the image is
smoothed before edge detection is
done. If the image is noisy, then
applying pre smoothing to the
image reduces the impact of the
noisy pixels on the edge detection
• Threshold: Consider each pixel,
and if the difference between this
pixel’s value and one of its
neighboring pixels is bigger than the
threshold, that pixel is considered as
a candidate for edge pixel. The
appropriate threshold value is
dependent on the value differences
of neighboring pixels along the
edges. The default is 18.
• Minimal Length Specifies the
minimum acceptable length of the
edge. Any edge lengths less than
this number are dropped. Minimal
length is measured in pixels. The
default is 3.

9
Segment Geometry
Settings
• Minimum Size. If this box is
checked, the number field next
to it is enabled. The minimum
size constraint is enforced using
the number specified in that
field.
• Compute Settings. Click this
button to change the default
values for Minimum Value
Difference and Threshold.

10
Segmentation
Parameters
• Minimal Value Difference
Enter the minimum value
difference between neighboring
segments. The default is 15.
• Variance Factor: by specifying
a larger variance factor, you can
reduce the number of segments
on non-homogeneous regions.
• Find Narrow Strips If this
checkbox is checked, strips of
one or two pixels wide are
extracted as separate
segments. Otherwise, narrow
strips are extracted only if their
width is larger than three pixels.

11
Esercizio 1 – Image
Segmentation
Immagine =
l2a_20170214t094559_10m_aoi
Output =
edge_01.img
segmentation_01.img

13
Segmentation
Immagine =
l2a_20170214t094559_10m_aoi
Output =
edge_02.img
segmentation_02.img

15
Segmentation
Immagine =
l2a_20170214t094559_10m_aoi
No Edge Detection
Output =
segmentation_03.img

17
Segmentation
Immagine =
l2a_20170214t094559_10m_aoi
No Edge Detection
Aumentare il Variance Factor a 4
Output =
segmentation_04.img

20
FLS Image Segmentation
▪ Zonal Input Image: a binary raster
that contains zones
▪ Pixel Segment Ratio: Specify
average number of pixels that each
output segment will contain.
▪ Relative Weights: The parameters
in the Relative Weights group
control the merge cost function.

21
Relative Weights:
Spectral: Specify how much weight to give the
spectral component of the segments. This is
measured as the mean of the DN values of the pixels
in the segments.
▪ Enter a high value to generate segments that
are more spectrally homogeneous.
▪ Enter a low value to generate segments that
are less spectrally homogeneous

22
Relative Weights:
Texture - Specify how much weight to give the
texture component of the segments. This is
measured as the standard deviation of the DN
values of the pixels in the segments.
are more texturally homogeneous.
are less texturally homogeneous.

23
Relative Weights:
Size - Specify how much weight to give the size
component of the segments. This is measured as the
number of pixels in the segment.
are more homogeneous in size.
are less homogeneous in size.

24
Relative Weights:
Shape - Specify how much weight to give the shape
component of the segments. This is a proprietary
measurement of the boundary complexity of the
segment.
are more homogeneous in shape (compactness
and smoothness).
are less homogeneous in shape (compactness
and smoothness).

25
Size Limits - Size Limits may be specified to
constrain the size of the segments. These
parameters are expressed in terms of pixel count.
▪ Minimum - Specify the minimum size of the
segments in pixels.
▪ Maximum - Specify the maximum size of the
segments in pixels.

26
Image Segmentation Tools in ERDAS: a
comparison
Image Segmentation
▪ An edge detection and region
growing algorithm
▪ Considers only spectral content of
the segments for merging decisions
▪ No direct control over
pixel:segment ratio
▪ Considers spectral content as well
as the segment's texture, size, and
shape for merging decisions
▪ Direct control over the
pixel:segment ratio
▪ Both minimum and maximum
segment size constraints

27
FLS Image Segmentation – Esercizio 1
▪ No Zonal Input Image
▪ Default Values
▪ Output:
fls_segmentation_01.img

28
▪ No Zonal Input Image
▪ Output:

29
▪ No Zonal Input Image:
▪ Output:

Image segmentation in erdas imagine

In this document

More Related Content

What's hot

Similar to Image segmentation in erdas imagine

More from Planetek Italia Srl

Recently uploaded

Image segmentation in erdas imagine