图像分割过分割和欠分割_使用图割的图像分割

图像分割过分割和欠分割

Image segmentation technology is an important research direction in the field of computer vision and an important part of image semantic understanding. Image segmentation refers to the process of dividing an image into several areas with similar properties. From a mathematical point of view, image segmentation is the process of dividing an image into disjoint areas.The area can be the foreground and background of the image or a single object. These areas can be constructed using features such as color, edges, or similarity of neighbors.

图像分割技术是计算机视觉领域的重要研究方向，也是图像语义理解的重要组成部分。 图像分割是指将图像划分为多个具有相似属性的区域的过程。 从数学的角度来看，图像分割是将图像划分为不相交的区域的过程，该区域可以是图像的前景和背景，也可以是单个对象。 可以使用诸如邻居的颜色，边缘或相似性之类的特征来构造这些区域。

Graph cutting algorithm is one of the classic algorithms of combinatorial graph theory. In recent years, many scholars have applied it to image and video segmentation and achieved good results. This article briefly introduces the graph cut algorithm and interactive image segmentation technology, as well as the application of graph cut algorithm in interactive image segmentation.

图切割算法是组合图理论的经典算法之一。 近年来，许多学者将其应用于图像和视频分割，并取得了良好的效果。 本文简要介绍了图割算法和交互式图像分割技术，以及图割算法在交互式图像分割中的应用。

基本概念 (Basic concept)

We use the theories and methods in the field of graph theory to map the image into a weighted undirected graph, treat the pixels as nodes, and regard the image segmentation problem as the vertex division problem of the graph, using the smallest cutting criterion to obtains the best segmentation of the image.

我们使用图论领域中的理论和方法，将图像映射成加权无向图，将像素作为节点，并将图像分割问题视为图的顶点划分问题，并使用最小的切割准则获得图像的最佳分割。

This type of method associates the problem of image segmentation with the problem of MIN-CUT . The usual approach is to map the image to be segmented into a weighted undirected graph G=(V, E), where , V is the set of vertices, and E is the set of edges.

这种类型的方法将图像分割问题与MIN-CUT问题联系起来。 通常的方法是将要分割的图像映射到加权无向图G =(V，E)，其中，V是顶点集，E是边缘集。

The edge formed by the connection of every two neighboring vertices are called n-links and the connection between each ordinary vertex and the two terminal vertices are called t-links.

每两个相邻顶点之间的连接所形成的边称为n-link ，每个普通顶点与两个终端顶点之间的连接称为t-link。

Each node corresponds to each pixel in the image, and each edge ∈ E connects a pair of adjacent pixels, and the weight of the edge is w(i,j) represents the non-negative similarity in gray, color or texture between adjacent pixels.

每个节点对应于图像中的每个像素，每个边缘∈E连接一对相邻的像素，并且边缘的权重为w(i，j)表示相邻像素之间在灰度，颜色或纹理上的非负相似度。

Boykov and Jolly originally proposed to compute the his