* MATLAB demo code for object matching of ICCV 2009 paper

Minsu Cho, Jungmin Lee, and Kyoung Mu Lee, 
Feature Correspondence and Deformable Object Matching via Agglomerative Correspondence Clustering, 
Proc. International Conference on Computer Vision (ICCV), 2009.

Please consider citing our work if you find this code useful in your research. 
Note that this is a simple MATLAB implementation without consideration for speed-up. 

written by Minsu Cho, 2009, Seoul National University, Korea
http://cv.snu.ac.kr/~minsucho

OVERVIEW

run_demo.m          : script for demo
config_file.m       : config file for demo
show_initialmatches : script for visualizing initial matches
show_results        : script for visualizing final results

For details of algorithm parameters, refer to our paper.

This code requires an initial match file for running.

Two exemplar data files for demo are included in this version,
1) match_All1+Magazines.mat               
2) match_OvoLeoSuchardB+OvoLeoSuchardC.mat
, each denoting initial matching between (image1)+(image2).
The images are borrowed from the ETHZ toys dataset, 
http://www.vision.ee.ethz.ch/~vferrari/datasets.html

Each file includes two kinds of data structure, viewInfo and mathInfo

viewInfo - view information of each image (1 or 2)
    viewInfo(1 or 2).img              = image data
    viewInfo(1 or 2).fileName         = file name
    viewInfo(1 or 2).feat(:,1:5)      = [ u, v, a, b, c ] (affine region parameters)
                                        a(x-u)(x-u)+2b(x-u)(y-v)+c(y-v)(y-v)=1
    viewInfo(1 or 2).affMatrix(:,1:9) = 3 by 3 affine trasformation 
                                        from the unit circle to the region
    viewInfo(1 or 2).desc(:,1:128)    = sift descriptor for each feat

matchInfo(:) - match information for each match
    matchInfo(:).match = [ row index at view(1).feat,  row index at view(2).feat ]
    matchInfo(:).dist  = dissimilarity between two features ( sift desc distance )

This data file was constructed based on the codes and binary files from Oxford Visual Geometry Group,
http://www.robots.ox.ac.uk/~vgg/research/affine/index.html
