This preview shows page 1. Sign up to view the full content.
Unformatted text preview: WRET2101
Representation and Description Representation & Description The results of segmentation is a set of regions. Regions have then to be represented and described. Representation Two main ways of representing a region:
1. external characteristics (its boundary): focus on shape internal characteristics (its internal pixels): focus on color, textures... 2. Description The next step: description E.g.: a region may be represented by its boundary, and its boundary described by some features such as length, regularity... Features should be insensitive to translation, rotation, and scaling. Both boundary and regional descriptors are often used together. Chain Codes Chain codes are used to represent a boundary by a connected sequence of line segments of specified length and direction Representation is based on 4 or 8 connectivity of segments. 4directional chain codes Chain Code Example
It may be useful to downsample the data before computing the chain code - to reduce the code dimension - to remove small detail along the boundary Representation of Boundaries The chain code of boundary depends on the starting point. Can remove the dependence from the starting point by redefining the starting point so that the resulting sequence of numbers forms an integer of minimum magnitude. Method 1 Chain Codes Traverse the boundary in clockwise (or ACW) direction and assign a direction to the segments connecting every pair of pixels. However, the resulting chain code is usually quite long and sensitive to noise effects. Method 2 Resample the boundary by selecting a larger grid spacing. As the boundary is traversed, assign a boundary point to the node of the grid to which it is closest. Method 3 Divide the boundary into segments of equal length (each segment having the same number of pixels). Join the endpoints of each segment by a straight line. The resulting code is 217005552 Boundary Descriptors Every image has object with certain borders/countours Chain code produce strings that describes about the shape of the object To describe the shape of the chain code, we need boundary descriptors, a mechanism to compare the shape of the chain code. Like measuring the distance of the strings produced by the chain code. Distance measuring technique Euclidean Euclidean Distance Definition: The straight line distance between two points. In a plane with p1 at (x1, y1) and p2 at (x2, y2), it is ((x1 x2) + (y1 y2)). Euclidean Distance Assume an image A is represented with a vector that contains information regarding image's shape. In a constant space, similarity of image A and image B, each with vector P and Q, and can be measured using the following Euclidean Distance: (i=1,N (PiQi)) Distance = 0 both images are the same The smaller the distance the more similar the image is with the image in query. Regional Descriptors
Topology is the study of the properties which are unaffected by any deformation (rubber-sheet distortion) - number of holes H - number of connected components C - Euler number, E=C-H Steps for Basic Image Retrieval based on Shape
1. 1. Edge detection Sobel Edge Detection 1. Describe the shape of an object Chain Code Search and Match Shape Using technique in measuring distance in string produced by (2) like Euclidean Distance Matlab Exercise Briefly describe the following matlab functions: bwdist(BW) bweuler(BW,n) ...
View Full Document
- Winter '09