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 > IT/计算机 > 计算机软件及应用 > An Efficient Method for Color Image Segmentation
侵权投诉

An Efficient Method for Color Image Segmentation

图像分割、彩色图像、meanshift

Algorithm for Adaptive Mean Shift Procedure :

Input : A positive integer k, n data points, Xi, where ª013ffº«104f2»

i=1,2,……,n, one of the data points y1, and a total number N of iterations.

Output: An appropriate local maximum (mode) of the density. 1.1.hi xi xi,k1

,where xi,k is the k-nearest neighbor of xi

2.2. j = 1

3.

3. while j < N do

¦n1§y1 x2

i·i 1hd2xig¨

¨h¸¸4. yi2=

i¦n1§y x2i·i 1hd2g¨

1i¨©h¸i¸¹

5. y1= y2

6. j = j + 17. End while8. Output y2

If hiis fixed in Procedure 1, then this is called the Mean Shift Procedure, implementing the main principle of mean-shift based clustering.

A. Graph Partitioning-Normalized Cuts

This method is used for solving the perceptual grouping problem in vision. This method doesn’t consider the local features and their consistencies in the image data. This criteria measures both the total dissimilarity between the different groups as well as the total similarity within the groups Consider a Graph G = (V, E, W), where V is the set of vertices, E is the set of edges and W is the weight between two nodes. Let x and y be two elements in V and they are connected through a edge in E and the weight between these nodes are represented in W. W is actually a weight matrix that stores the dissimilarity between the vertices x and y. this graph can be partitioned into two disjoint sets A and B. Set B=V-A.ieA B=V and A B I by simply removing the edges connecting these two sets. In graph theory, a cut is defined as the partitioning of the graph into different groups having strong interior links and weaker exterior links. Fig 2.1(a) represents a weighted graph and the matrix shown in 2.1(b) describes the weight matrix. A cut of a graph G is the set of edges S such that removal of S from Gdisconnects G. A minimum cut is the cut of minimum weight. The degree of dissimilarity between A and B is computed as the total weight of the edges that have been removed. The weight of cut(A,B) is given as Cut(A,B) =

¦

x A,y B

w(x,y)

An Efficient Method for Color Image Segmentation

(6)

978-1-4673-0671-3/11/$26.00©2011 IEEE «

««

34067»»

«ff601»«»¬f2710¼»

Figure 2.1 (a) Weighted Graph (b) Weight Matrix

The optimal bi-partitioning of a graph is the one that

minimizes this cut value. But the number of such partitions is exponential. This problem is solved by normalized cut method [4]. Normalized cuts method computes the cut cost as a fraction of the total edge connections to all the nodes in the graph. This disassociation is called the normalized cut. In minimum cut method, it only computes value of total edge weight connecting the two partitions. The Normalized cut is given as : Ncut (A,B) =

cut(A,B)cut(A,B)

assoc(A,V)

assoc(B,V)

(7) where assoc (A,V) =

¦

x A,y V

w(x,y) is the total

connection from nodes in A to all nodes in the graph, and assoc(B,V) is similarly defined. With this definition of the disassociations between the groups, the cut that partitions out small isolated points will no longer have small Ncut value. Since the cut value is almost certainly be a large percentage of the total connection from that small set to all other nodes. The total normalized association within groups for a given partition: Nassoc (A,B) =

assoc(A,A)assoc(B,B)

assoc(A,V)

assoc(B,V)

(8) III PROPOSED APPROACH - ADAPTIVE MEAN SHIFT

ALGORITHM WITH NORMALIZED CUTS Algorithm:

In the first stage, the color image is segmented using adaptive mean shift algorithm. In mean-shift algorithm, if we use constant window size used for all the feature point windows, that is hi–h=constant, the algorithm is called the constant bandwidth mean shift. If we use adaptive value as the window size at each feature point xiwith hi= h(xi), the algorithm is called the adaptive mean shift(AMS). Adaptive Mean Shift clustering has been produces better results than the normal mean-shift algorithm especially in high dimensional feature space. Many methods are used to find the adaptive window size for the Adaptive Mean Shift algorithm [18] [24]. An

311

IEEE-ICoAC 2011

上一页第3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