170 likes | 322 Views
Data Structures For Image Analysis. Levels of image data representation Traditional image data structures Hierarchical data structures. Levels Of Image Data Representation. Computer visual perception Determine the relation b/w input image and models of real world
E N D
Data Structures For Image Analysis Levels of image data representation Traditional image data structures Hierarchical data structures
Levels Of Image Data Representation • Computer visual perception • Determine the relation b/w input image and models of real world • Iconic image – original image data • Segmented image – ROI in groups • Geometric representation – higher level of knowledge, such as shapes, etc. • Relational model – relationships among higher level abstraction Image-based Digital Image Processing
Traditional Image Data Structures • Matrices or N-dimensional arrays • Chains – describing object borders • Topological data structures – graphs, maps • Relational structures Digital Image Processing
Matrices • Low-level image data representation • Depict spatial relations – neighborhood, etc. • Grid – rectangular, hexagonal grids • Pixel coordinates • Brightness – intensity, gray level, color Digital Image Processing
Matrices (II) • Binary image (0/1), multi-spectral image (gray-scaled, color), hierarchical image data structure (LOD: level of detail, varied resolutions) • Global information • Histogram – probabilistic density of a phenomenon • Co-occurrence matrix – measures in terms of brightness Digital Image Processing
Co-occurrence Matrix The diagonal elements correspond to the histogram! Digital Image Processing
Chains • Chains are used for the description of object borders in computer vision • Chains are composed of symbols in sequence – useful for syntactic pattern recognition • Chain codes (aks: Freeman codes) • Run length coding Digital Image Processing
Chain Codes 0007766555555670000006444444442221111112234445652211 00077665555556600000006444444442221111112234445652211 (X) Digital Image Processing
Run Length Coding ((11144)(214)(52355)) Digital Image Processing
Topological Data Structures • Describe image as set of elements and their relations • Graph: G=(V,E); V denotes the set of nodes and E represents the set of edges • Evaluated graph (or weighted graph) • Region adjacency graph Digital Image Processing
Region Adjacency Graph (RAG) • Nodes represent region; edges or arcs represent connectivity • Nodes of degree 1 are cavities or holes • Edges can be used to describe relations • RAG can be created from a quadtree representation or from tracing the borders of all regions in the region map (a result of segmentation) Digital Image Processing
Region Merging Phenomenon Region merging may create holes Digital Image Processing
Relational structure Digital Image Processing
Pyramids • M-pyramid (matrix-pyramid) – a sequence of images in reducing resolutions of the original image • Disadvantage: Only one image in certain resolution is available at a time • T-pyramid (tree-pyramid) – use the tree structures to represent M-pyramid Digital Image Processing
T-Pyramids Digital Image Processing
Quadtree Similar to pyramid hierarchical representations. T-pyramids are balanced; the quadtree representation is not. Digital Image Processing