边界跟踪算法 输入是一幅图像，输出时边界坐标。基于MATLAB的

图像分割时，经常需要实现对边界的跟踪，从而实现对区域的提取。边界跟踪算法输入时一幅二值图像（即边界图像），实现对边界的定位，从而在得到边界的基础上提取相应区域。输出是边界坐标。

function C = contour_following（BW）
% CONTOUR_FOLLOWING takes a binary array and returns the sorted row and
% column coordinates of contour pixels.
%
% C = CONTOUR_FOLLOWING（BW） takes BW as an input. BW is a binary array
% containing the image of an object （‘1‘: foreground ‘0‘: background）. It
% returns a circular list （N x 2 C（1:）=C（end:）） of the 
% （rowcolumn）-coordinates of the object‘s contour in the order of 
% appearence （This function was inspired from the freeman contour coding 
% algorithm）.
%
% Note: 
% - if the object is less than 3 pixels CONTOUR_FOLLOWING sends back [0 0].
% - the algorithm is quite robust: the object can have holes and can also
% be only one pixel thick in some parts （in this case some coordinates
% pair will appear two times: they are counted “way and back“）.


[mn]=size（BW）;                                                            % getting the image height and width

Itemp=zeros（m+2n+2）;                                                      % we create a ‘0‘ frame around the image to avoid border problems
Itemp（2:（m+1）2:（n+1））=BW;
BW=Itemp;

BW = BW - imerode（BW[0 1 0 ; 1 1 1 ; 0 1 0]）;                             % gets the contour by substracting the erosion to the image
BW = bwmorph（BW‘thin‘Inf）;                                               % to be sure to have strictly 8-connected contour

if （sum（sum（BW））<3）                                                       % we consider that less than 3 pixels cannot make a contour
    C=[0 0]; 
    return; 
end;

[rowcol]=find（BW1）;                                                      % takes the first encountered ‘1‘ pixel as the starting point of the contour

MAJ=[6 6 0 0 2 2 4 4];                                                     % variable initialization
C=[0 0 ; 0 0];
k=0;
ended=0;
direction=4;

while（ended==0）
    k=k+1;
    found_next=0;  
    
    while（found_next==0）
        switch mod（direction8）
            case 0
                if （BW（row col+1）==1）
                    row=row;
                    col=col+1;
                    C（k:）=[row col];
                    found_next=1;
                end;
            case 1;
                if （BW（row+1 col+1）==1）
                    row=row+1;
                    col=col+1;
                    C（k:）=[row col];
                    found_next=1;
                end;
            case 2;
                if （BW（row+1 col）==1）
                    row=row+1;
                    col=col;
                    C（k:）=[row col];
                    found_next=1;
                end;
            case 3;
                if （BW（row+1 col-1）==1）
                    row=row+1;
                    col=col-1;
                    C（k:）=[row col];
                    found_next=1;
                end;
            case 4;
                if （BW（row col-1）==1）
                    row=row;
                    col=col-1;
                    C（k:）=[row col];
                    found_next=1;
      

 属性            大小     日期    时间   名称
----------- ---------  ---------- -----  ----

     文件       4208  2010-05-27 19:16  55593365contour_following\contour_following.m

     文件      36054  2007-05-10 13:36  55593365contour_following\test1.bmp

     文件      36054  2007-05-10 13:37  55593365contour_following\test2.bmp

     文件      36054  2007-05-10 13:37  55593365contour_following\test3.bmp

     文件      36054  2007-05-10 13:37  55593365contour_following\test4.bmp

     文件      36054  2007-05-10 13:38  55593365contour_following\test5.bmp

     文件       1103  2007-05-10 13:39  55593365contour_following\Testscript.m

     文件         61  2010-05-27 19:20  55593365contour_following\程序说明.txt

     目录          0  2010-05-27 18:50  55593365contour_following

----------- ---------  ---------- -----  ----

               185642                    9