根据某些条件查找连接组件标签

Question

可以使用哪些lib来查找python中连接的组件标签，以便邻居像素应该满足某些条件，例如（left_up / current＆lt; = 3.0等）。

哪种算法最适合用于大型图像连接组件标记。

Answer 1

这是代码

def connected_components(self,SWTImage):
    width=SWTImage.shape[1]
    heigth=SWTImage.shape[0]
    G=nx.Graph()
    row_loc=np.zeros(SWTImage.shape[0]*SWTImage.shape[1])

    for y in xrange(1,heigth):
        for x in xrange(1,width):
            currentPix=SWTImage.item(y,x)
            if currentPix>0:
                left=x-1
                up=y-1
                rigth=x+1

                leftPix=SWTImage.item(y,left)
                if(leftPix>0 and (leftPix/currentPix<=3.0 or currentPix/leftPix<=3.0)):
                    G.add_edge(y*width+x,y*width+left)
                    row_loc.itemset((y*width+left),y)

                leftUpPix=SWTImage.item(up,left)
                if(leftUpPix>0 and (leftUpPix/currentPix<=3.0 or currentPix/leftUpPix<=3.0)):
                    G.add_edge(y*width+x,up*width+left)
                    row_loc.itemset((up*width+left),up)

                upPix=SWTImage.item(up,x)
                if(upPix>0 and (upPix/currentPix<=3.0 or currentPix/upPix<=3.0)):
                    G.add_edge(y*width+x,up*width+x)
                    row_loc.itemset((up*width+x),up)

                if(rigth<width):
                    rightUpPix=SWTImage.item(up,rigth)
                    if (rightUpPix>0 and (rightUpPix/currentPix<=3.0 or currentPix/rightUpPix<=3.0)):
                        G.add_edge(y*width+x,up*width+rigth)
                        row_loc.itemset(up*width+rigth,up)

                row_loc.itemset((y*width+x),y)

    components=nx.algorithms.components.connected_components(G)
    return components,row_loc

我使用DFS查找连接组件列表。

Answer 2

这样做的经典方法是使用here数据结构描述disjoint set的双遍算法。它可以很容易地修改为适用于任意条件。