Segmentation
Read more
about Segmentation in SPRING
Image Segmentation:
-
selecton the"Control Panel" an image and click on Image - Segmentation... or DTM - Segmentation... on the main menu;
-
select the image(s) that you want
to use for segmentation;
-
choose a segmentation Method:
Region Growing , Region Growing Multi, Region Growing Baatz or Basin Detection;
If in this case the choice of the segmentation method is Region Growing:
-
type in a value, integer and
greater than zero, that will be used as Similarity limit;
-
type in a value, integer and
greater than zero, that will be used as a minimum size of Area,
in pixels, that represents the segmented region;
-
choose
a Band of Exclusion
in case you do not want the image region, defined by this band, to be
considered during the classification process. The band of exclusion is
used for defining regions where you do not want to segment the chosen
image.
If in this case the choice of the segmentation method is Region Growing Multi:
-
type in
a value, integer and
greater than zero, that will be used as Similarity limit. Use lower values to merge only segments that are more similar -
Higher values will allow more segments to be merged - valid values range: positive values and greater than zero;
-
type in
a value, integer and
greater than zero, that will be used as a minimum size of Area,
in pixels, that represents the the segmented region;
-
if the Normalize button is enabled, the provided Similarity value will be normalized. If the value supplied is
a normal value, the button must be disabled. The default value is triggered. This also applies to images with
very low pixel values. This function is only enabled for this method.
-
type in
Optionally, click Advanced Operations to change the default values of the performance-related advanced parameters (see about operation of Advanced Operations);
If in this case the choice of the segmentation method is Region Growing Baatz:
-
type in
a value, integer and
greater than zero, that will be used as Similarity limit. Use lower values to merge only segments that are more similar -
Higher values will allow more segments to be merged - valid values range: between [0,1], positive values and greater than zero;
-
type in
a value, integer and
greater than zero, that will be used as a minimum size of Area,
in pixels, that represents the the segmented region;
-
type in
a value for Color Weight: The weight given to the color component, default: 0.5, valid range: [0,1];
-
type in
a value for Compactness: The weight given to the compaction component, standard: 0.5, valid range: [0,1];
-
type in
Optionally, click Advanced Operations to change the default values of the performance-related advanced parameters (see about operation of Advanced Operations);
Note: For more information about the Baatz method based on the article "Multiresolution segmentation: an optimization approach for high quality multi-scale segmentation"
by Baatz and Schape, 2000; click here.
If in this case the choice of the segmentation method is Basin Detection:
-
type in
a value, integer and
greater than zero, that will be used as initial threshold value for ND.
Note: Basin detection segmentation is done on an image resulting from edge extraction.
Edge extraction is performed by an edge detection algorithm, that is, by the Sobel filter.
This algorithm considers the gray level gradients of the original image to generate a gradient image
or edge intensity image.
-
type in a name for an IL that will
contain the Segmented Image to be created;
-
in
Arc Smoothing, choose Yes if you want to smooth
the borders or No if you don't;
-
click on Bounding
Box…
to define an area smaller than the project’s area, in case you
want to perform a segmentation only in part of the image (see about the bounding box operation);
-
click on Apply to perform
the segmentation.
Notes:
1. The Similarity measure is based on the Euclidean distance between the
average values of gray levels of each region. Therefore, two regions are
considered different if the distance between their averages is greater than
the Similarity limit chosen.
2. Regions with areas smaller than the minimum chosen are absorbed by
adjacent regions more similar to them.
|