Steve on Image Processing

June 13th, 2007

I've about run out of things to say for now about connected component labeling algorithms. ("Hurray!" shouts the audience.)

For reference, here's a summary of all the posts in the series:

• Part 1 - Introduction
• Part 2 - Connectivity definitions
• Part 3 - Adjacency matrices and dmperm
• Part 4 - Flood fill
• Part 5 - Two-pass algorithm with equivalence class resolution
• Part 6 - Union-find
• Part 7 - Run-length encoding

16:00 UTC | Posted in Connected components | Permalink | 2 Comments »

You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.

2 Responses to “Connected component labeling – Wrapping up”

1. Ram replied on February 16th, 2008 at 00:46 UTC :
   

   hi steve,

   Is there a way of establishing custom connectivity
   such as conn=[00100;
   01110;
   11111;
   01110;
   00100];?
   If i am not wrong bwlabel allows only scalars 4 and 8.

2. Steve replied on February 16th, 2008 at 08:14 UTC :
   

   Ram—Use bwlabeln, which allows custom connectivity matrix. However, your example is not a valid connectivity matrix. The Image Processing Toolbox connectivity definitions do not allow you to say that a pixel is directly adjacent to another pixel that’s two steps away, which is what your matrix implies.

These postings are the author's and don't necessarily represent the opinions of The MathWorks.