SHAPE ANALYSIS INEL 6088 Computer Vision Refs.: ch. 6, Davies; Ch. - - PowerPoint PPT Presentation
SHAPE ANALYSIS INEL 6088 Computer Vision Refs.: ch. 6, Davies; Ch. 2 Jain et al. TOPICS Connected components CC Labeling Image moments Geometrical properties: position and orientation Region-boundary following Algorithm
INEL 6088 Computer Vision
Refs.: ch. 6, Davies; Ch. 2 Jain et al.
Distance metrics
q − − d4 = |xi − xj| + |yi − yj| = ( ) dE = q (xi − xj)2 + (yi − yj)2 | − | | − | d8 = max (|xi − xj|, |yi − yj|)
Euclidean City-block Chessboard
Distance transform: provides the minimum distance from a pixel of an object to the background. To compute, use where m is the iteration number for all pixels [u,v] such that d([u,v],[i,j])=1. Use 4-neighbors.
skeleton, = medial axis = set of pixels that are at a locally- maximum distance from the background A pixel [i,j] is locally maximum if its distance to S is not less than that of all its neighboring pixels [u,v]. Here we are using 4- neighbors.
Skeleton S*
skeleton, = medial axis = set of pixels that are at a locally- maximum distance from the background A pixel [i,j] is locally maximum if its distance to S is not less than that of all its neighboring pixels [u,v]. Here we are using 4- neighbors.
THINNING
THINNING
examine its 3x3 pixel- neighborhood
the boundary until the region have been reduced to thin-lines
maintain 8-connectivity
line
EXPANDING AND SHRINKING
Expanding: Change pixel from 0 to 1 if any neighbors of the pixel are 1. Shrinking: Change a pixel from 1 to 0 if any neighbors of a pixel are 0. Expanding followed by shrinking can be used to fill holes. Shrinking followed by expanding can remove isolated noise pixels.