计算机视觉概述mean-shift-Ncuts.pptVIP

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * degree (or valency) of a vertex of a graph is the number of edges incident to the vertex, with loops counted twice * degree (or valency) of a vertex of a graph is the number of edges incident to the vertex, with loops counted twice 基于图的图像分割 只能得到孤立的解 主要内容 图像分割的历史 基于图的图像分割 算法介绍 研究现状 下一步研究 算法流程 Assign W(i,j) Solve for minimumpenalty Cut into 2 Subdivide? Subdivide? Yes Yes No No 输入:图像 输出:分割块 每次迭代分割成两块 权重函数W(i,j) 颜色可能会包含比较多的噪声! 可以使用亮度和位置 亮度特征项 位置项 目标函数的最小化 与子集A连接的所有边的权重和 与割连接的所有边的权重和 通过对目标函数的最小化进行求解 Partition A Partition B cut 通过对目标函数的最小化进行求解 W(N x N) : weights associated with edges D (N x N) : diagonal matrix with summation of all edge weights for the i-th pixel N : number of pixels in the image 求解归一化的拉普拉斯特征方程 O(N^3) complexity in general O(N^(3/2)) complexity in practice a) Sparse local weights, b) Only need first few eigenvectors, c) Low precision (N) : eigenvalues (N x N) : eigenvectors are real-valued partition indicator Laplacian Matrix Definition degree matrix is a diagonal matrix which contains information about the degree of each vertex. adjacency matrix is a means of representing which vertices of a graph are adjacent to which other vertices Laplacian matrix is the difference of the degree matrix and the adjacency matrix of the graph. Laplacian Matrix Definition Laplacian matrix is the difference of the degree matrix and the adjacency matrix of the graph. spectral graph theory Laplacian matrix is a matrix representation of a graph 可以根据第二大的特征向量对图像进一步的分割 细分? Threshold ? Yes – stop here No – continue to subdivide 主要内容 图像分割的历史 基于图的图像分割 算法介绍 研究现状 下一步研究 研究现状: 边权重函数 Probability of boundary on line from to Advancements in edge detection No Boundary Boundary 研究现状: 标准测试数据库BSDS Berkeley Segmentation Dataset (BSDS) 研究现状: Best Technique Normalized Cuts is base technique for best low lev