hausdorff distance and it's application to tile images classification

前段时间有会友发过求瓷砖图片分类的帖子，我看这是一个演示hausdorf距离的好例子，所以就发到这里来了。

hausdorf 距离的介绍请参考如下文献：

[1]   D. P. Huttenlocher, G. A. Klanderman, and W.  J. 

      Rucklidge, “Comparing  images using the Hausdorff 

      distance”,  IEEE  Trans. PAMI,  vol.  15, pp.  850- 

      863, 1993.

[2]  Marie-Pierre Dubisson and A.K.Jain. A modified hausdorff distance for

     object matching

这里用的是文献[2] modified hausdorff distance.

核心函数：

function d = m_hd(im1, im2)
%Effect: compute the modified haussdorff distance between 2 bw image 'im1'
%        and 'im2'
%inputs:
%im1, im2: bw image with the size
%outputs:
%d: the modified haussdorff distance between 'im1' & 'im2'
%Author: Su dongcai at 2012/1/11
%Email: suntree4152@gmail.com, qq:272973536
%reference:
%[1]   D. P. Huttenlocher, G. A. Klanderman, and W.  J. 
%      Rucklidge, “Comparing  images using the Hausdorff 
%      distance”,  IEEE  Trans. PAMI,  vol.  15, pp.  850- 
%      863, 1993.
%[2]  Marie-Pierre Dubisson and A.K.Jain. A modified hausdorff distance for
%     object matching

%0. check inputs:
im_sz1 = size(im1); im_sz2 = size(im2);
if(sum(abs(im_sz1-im_sz2))~=0)
    error('im1 and im2 must with the same size');
end
dist_im1 = bwdist(im1); dist_im2 = bwdist(im2);
pidx_im1 = find(im1); pidx_im2 = find(im2);
%eq.(6) in [2]
d12 = mean(dist_im1(pidx_im2));
d21 = mean(dist_im2(pidx_im1));
%eq.(8) in [2]
d = max(d12, d21);

结果演示：

splice_alltiles('.\tileImgs', '.\', 'splicedTiles_sort',100, 10, 10, [1:100]);

得到原始未分类图片：

[sort_idx, num_indv] = sortImgs('.\tileImgs', 100, 1.5);

splice_alltiles('.\tileImgs', '.\', 'splicedTiles_sort',100, 10, 10, sort_idx);

得到分类图片：

[local]2[/local]

num_indv =

    21    14    12    11     2     2    10    12     4     1     1     3     3     1     3

表示按列优先顺序（从上往下，从左往右），前21块砖为1类， 紧接着的14块砖为另一类，以此类推。