Argthtjtr

I am trying to find a flood filling function that can give me as output a list with the coordinates (min and max of the x,y axis) of the clusters ('connected'pixels).
I have been given the following implementation but it seems to be the most time consuming part of my algorithm, and I wanna speed it up, as in the end it has to run over some hundreds of images.

    def floodfill8(self,coord,i,j,imin,imax,jmin,jmax,area=0):

    r"""determining cluster size and coordinates by recursive flood filling    

    """



    if coord[i,j]==1:

        area+=1

        imax=np.max([imax,i])

        imin=np.min([imin,i-1])

        jmax=np.max([jmax,j])

        jmin=np.min([jmin,j-1])

        coord[i,j]=0

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i,j+1,imin,imax,jmin,jmax,area)

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i+1,j,imin,imax,jmin,jmax,area)

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i,j-1,imin,imax,jmin,jmax,area)

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i-1,j,imin,imax,jmin,jmax,area)

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i+1,j+1,imin,imax,jmin,jmax,area)

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i+1,j-1,imin,imax,jmin,jmax,area)

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i-1,j+1,imin,imax,jmin,jmax,area)

        area,coord,imin,imax,jmin,jmax=self.floodfill8(coord,i-1,j-1,imin,imax,jmin,jmax,area)

    return area, coord, imin, imax, jmin, jmax



    def findcluster(self,coord):

    r"""Cluster finding algorithm, finds connected groups of 1's in coord

    ---------

    Input:

        coord      array of 0's and 1's

    ---------

    Output:

        clusters   list of clusters with elements [xmin,xmax,ymin,ymax,size]"""

    #add fixed boundaries

    newcoord=np.zeros((coord.shape[0]+2,coord.shape[1]+2)) 

    newcoord[1:-1,1:-1]=coord

    clusters=

    for i in np.arange(coord.shape[0]):

        for j in np.arange(coord.shape[1]):

            if newcoord[i+1,j+1]==1: 

                size,newcoord,imin,imax,jmin,jmax=self.floodfill8(newcoord,i+1,j+1,imin=i,imax=i,jmin=j,jmax=j)



                clusters.append([imin,imax,jmin,jmax,size])

    clusters=np.array(clusters)

    if len(clusters)>1:

        ind=np.argsort(clusters[:,-1])

        clusters=clusters[ind]

    return clusters

Note that this recursive flood filling is called in a loop over all the pixels in the image.
I really couldn't find anything already implemented, a library or something, and I am still quite a noob in this ,image processing, field, but I believe that it should be something optimized in some library.
I was wondering whether anyone has any information of a flood filling function or the relevant library, that can also give me this output.
The reason I want this output is twofold:
1.So that I can find the position of the clusters and remove them (the rectangle surrounding them) from the picture so that I do background correction.
2.So that draw a (red)-rectangular around those clusters that have the wished contrast and the wished size.
Hope my question is clear! If not I will be happy to get some feedback and explain the unclear parts.
Also if you have any advice in order to speed it up it is most welcome!
Thank you!