Segmentation Flashcards
segmentation definition
partitioning of image into regions-of-interest
segmentation methods
– by intensity
– by region
– by boundary
– by morphology
segmentation depending on intensity
● easy
● widely used
● Histogram based
● noise prone
● no connectivity
Segmentation by boundaries can be achieved by
– look for sharp changes in intensity
– can be achieved by spatial filtering
sobel filter which type of segmentation
Segmentation by boundaries
Segmentation by region growth
– start with seed
– check intensity in neighborhood
– if intensity within window, set to 1
– iterate until no change
Morphological operations analyze morphology of image structures based on
set theory and topology
Morphological operations extract which image information
– shape
– size
– connectivity
– number
– boundary
Morphological operations are mostly performed on
binary images
Structuring elements (SE)
● small bitmask to probe the image
● scan origin of SE over image
● check overlap between SE and image
● set pixel(s) to zero (or one)
Dilation
expand region
like convolution
fills holes
erosion
shrink region
like deconvolution
eliminates details
opening
first erosion, then dilation
Closing
first dilation, then erosion
Segmentation by morphology
uses opening filter SE
