Image Segmentation & Template Matching Multimedia Signal Processing lecture on Petri Hirvonen
Image Segmentation
Terminology Image processing tools Examples Details of the Assignment Tracking Rolling Leukocytes With Shape and Size Constrained Active Contours Image segmentation based on maximum-likelihood estimation and optimum entropy-distribution (MLE–OED)
Image segmentation problem is basically one of psychophysical perception, and therefore not susceptible to a purely analytical solution.
Motivation: Image content representation Requirements: object definition & extraction Mathematical morphology is very useful for analyzing shapes in images. Basic tools: dilation A+B and erosion A – B Application: boundary detection Internal boundary: A - (A – B) External boundary: (A+B) - A Morphological gradient: (A+B) - (A – B) Assignment: object edges A - (A – B) (A+B) - A (A+B) - (A – B)
Dilation: Replace every point (x,y) in A with a copy of B centered at B(0,0) The result D is the union of all translations. Erosion: The resulting set of points E consists of all points for which B is in A AADDEE BB Structuring element, kernel = B Minkowski addition / subtraction
Image information
Segmenting SEM-images
Dilation of the thresholded block contains the thresholded gradient completely at the optimal threshold. (k and p are thresholds, D is dilation with a structuring element of radius r)
Information & colour
Nobuyuki Otsu, A Threshold Selection Method from Gray-Level Histograms, 1979 For bimodal distributions minimized maximized Histogram-based thresholding Otsu’s method
Probability of intensity k Mean of k
Hough transformRegion Of Interest Histogram
Length and width are the perpendicular distances on the original (thresholded) target area. Perimeter is computed by the Chain Code algorithm.
Template Matching
We have first created a DATABASE that contains the elements in the table. FOR-loop is executed for all templates Font_images{index} And the result is visualized in colours: Scale ? Rotation ?
Object perimeter