文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 高等教育 > 理学 > A Survey of Shape Feature Extraction Techniques
侵权投诉

A Survey of Shape Feature Extraction Techniques

A Survey of Shape Feature Extraction Techniques

60 Pattern Recognition Techniques, Technology and Applications To represent the AGR image, in [24], quad-tree method is applied. Each node in the quad-tree covers a square region of the bitmap. The level of the node in the quad-tree determines the size of the square. The internal nodes (shown by gray circles) represent “partially covered” regions; the leaf nodes shown by white boxes represent regions with all 0s while the leaf nodes shown by black boxes represent regions with all 1s. The “all 1s” regions are used to represent the shape, Figure 15(b). Each rectangle can be described by 3 numbers: its center C = (Cx,Cy) and its size (i.e. side length) S. So each shape can be mapped to a point in 3n-dimensional space (n is the number of the rectangles occupied by the shape region).

A Survey of Shape Feature Extraction Techniques

(a) (b)

Fig. 15. Adaptive resolution representations (a) Adaptive Grid Resolution (AGR) image; (b) quad-tree decomposition of AGR.

Due to the fact that the normalization before computing AGR, AGR representation is invariant under rotation, scaling and translation. It is also computationally simple.

5.2 Bounding box

Bounding box computes homeomorphisms between 2D lattices and its shapes. Unlike many other methods, this mapping is not restricted to simply connected shapes but applies to arbitrary topologies [25].

To make bounding box representation invariant to rotation, a shape should be normalized by the same method as for AGR (Subsection 5.1) before further computation. After the normalization, a shape S is a set of L pixels, S = {pk ∈ R2|k = 1, 2,… ,L} and also write |S| = L. The minimum bounding rectangle or bounding box of S is denoted by B(S); its width and height, are called w and h, respectively.

Figure 16 shows the algorithm flowchart based on bounding box that divides a shape S into m(row) × n (column) parts. The output B is a set of bounding boxes.

An illustration of this procedure and its result is shown in Figure 17.

To represent each bounding box, one method is that partial points of the set of bounding boxes are sampled. Figure 18 shows an example.

If v = (vx, vy)T denotes the location of the bottom left corner of the initial bounding box of S, and denotes the center of sample box Bij

A Survey of Shape Feature Extraction Techniques

, then the coordinates

上一页第18页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email