Windows下 Robhess SIFT源碼配置

本文轉載自查看原文 2017-11-01 22:04 1401 圖像處理

Robhess OpenSIFT 源碼下載：傳送門

為了進一步學習SIFT，選擇論文就着代碼看，在VS2013、OpenCV2.4.13下新建項目，跑一跑經典之作。由於將代碼和Opencv配置好后還會有些錯誤提示，所以下面是代碼的一些改動之處。（試了下其實還是ubuntu下更方便，因為有許多參數或者命令是linux下的，當然windows下可以進行一些修改后利用）。

大前提：opencv配置好。剩下的都可以通過修改來搞定。

首先看看解壓后的文件，我們只需要頭文件和源文件：

頭文件：6個源文件：10個

注意直接運行肯定不行，因為有幾個文件比較特殊，下面對所有頭、源文件進行解釋：

1）imgfeatures.c 和 imgfeatures.h:

有SIFT特征點結構struct feature的定義，除此之外還有一些特征點的導入導出以及特征點繪制函數的聲明，對應的imgfeatures.c文件中是特征點的導入導出以及特征點繪制函數的實現。

2）utils.c 和 utils.h:

這兩個文件中是一些圖像基本操作的函數，包括：獲取某位置的像素點，設置某位置的像素點（8位，32位和64位），計算兩點之間的距離的平方，在圖片某一點畫一個“X”，將兩張圖片合成為一個(在特征匹配中用到)，高是二者之和，寬是二者的較大者。

3）minpq.c 和 minpq.h：

這兩個文件中實現了最小優先級隊列(Minimizing Priority Queue)，也就是小頂堆，在k-d樹的建立和搜索過程中要用到。

4）kdtree.c 和 kdtree.h：

這兩個文件中實現了k-d樹的建立以及用BBF(Best Bin First)算法搜索匹配點的函數。如果需要對兩個圖片中的特征點進行匹配，就要用到這兩個文件。

5）xform.c 和 xform.h:

這兩個文件中實現了RANSAC算法(RANdom SAmple Consensus 隨機抽樣一致)。RANSAC算法可用來篩選兩個圖像間的SIFT特征匹配並計算變換矩陣。可以單利用RANSAC算法篩選兩個圖像間的SIFT特征匹配，以得到更好的匹配結果，很經典的算法，值得學習。

6）sift.c 和 sift.h:

論文里最主要的內容在此，里面的內容就是兩個特征點檢測函數sift_features()和 _sift_features()，sift_features()是用默認參數進行特征點檢測， _sift_features()允許用戶輸入各種檢測參數，其實sift_features()中也是再次調用_sift_features()函數。所以只需提供原圖像和存儲特征點的數組以及其他一些檢測參數，然后調用sift_features()或 _sift_features()就可完成SIFT特征點檢測。

7）siftfeat.c :含有main函數，用來實現特征點的檢測，返回特征點數目和標記特征點的圖像。（主要用到）

8）match.c : 含有main函數, 檢測兩張圖中的sift特征點，然后找到特征點的匹配對。（主要用到）

9）match_num.c : 含有main函數，檢測sift特征點，但是用到了linux下的多線程編程，所以這里暫時不做討論。

10）dspfeat.c : 含有main函數，可以從預先保存的特征點文件中讀取特征點並顯示在圖片上。

一. 修改代碼

第一步：將代碼中所有頭文件和源文件中的聲明改一下：

修改前：

#include <cv.h>
#include <cxcore.h>
#include <highgui.h>

修改后:

#include <opencv/cv.h>
#include <opencv/cxcore.h>
#include <opencv/highgui.h>

修改原因：因為直接利用找不到opencv路徑，所以調整路徑。

第二步:修改源文件代碼：

1. sift.c:

將函數 static IplImage*** build_gauss_pyr( IplImage* base, int octvs, int intvls, double sigma ) 中的代碼進行改動：

修改前：

  const int _intvls = intvls;
  double sig[_intvls+3], sig_total, sig_prev, k;

修改后：

  const int _intvls = intvls;
  double *sig = (double*)malloc(sizeof(double)*(_intvls+3)); 
  double sig_total, sig_prev, k;
  ...
  free(sig);   //子函數返回前釋放內存

修改原因：

源代碼中用變量_intvls+3作為數組的長度，但是VC的編譯器不是GCC，它不允許這樣做。DEV-C++使用的編譯器是GCC，它允許使用變量作為數組的長度定義數組。

當然利用C++中的 new或者vector也可以。參考 new、delete與malloc、free區別。

2. utils.c

首先刪除這兩行：

#include <gdk/gdk.h>
#include <gtk/gtk.h>

為啥刪掉？

gtk是一個功能強大、設計靈活的一個通用圖形庫，是GNU/Linux下開發圖形界面的應用程序的主流開發工具之一，GTK+也有Windows版本和Mac OS X版。在作者的源碼中gtk用來調整窗口來顯示圖像，因為我懶於裝gtk，所以直接利用opencv進行顯示，所以這里需要修改一些opencv的東西。

將函數進行改動：

修改前：

 1 void display_big_img( IplImage* img, char* title )
 2 {
 3   IplImage* small;
 4   GdkScreen* scr;
 5   int scr_width, scr_height;
 6   double img_aspect, scr_aspect, scale;
 7 
 8   /* determine screen size to see if image fits on screen */
 9   gdk_init( NULL, NULL );
10   scr = gdk_screen_get_default();
11   scr_width = gdk_screen_get_width( scr );
12   scr_height = gdk_screen_get_height( scr );
13 
14   if( img->width >= 0.90 * scr_width  ||  img->height >= 0.90 * scr_height )
15     {
16       img_aspect = (double)(img->width) / img->height;
17       scr_aspect = (double)(scr_width) / scr_height;
18       
19       if( img_aspect > scr_aspect )
20     scale = 0.90 * scr_width / img->width;
21       else
22     scale = 0.90 * scr_height / img->height;
23 
24       small = cvCreateImage( cvSize( img->width * scale, img->height * scale ),
25                  img->depth, img->nChannels );
26       cvResize( img, small, CV_INTER_AREA );
27     }
28   else
29     small = cvCloneImage( img );
30   
31   cvNamedWindow( title, 1 );
32   cvShowImage( title, small );
33   cvReleaseImage( &small );
34 }

View Code

修改后：

void display_big_img(IplImage* img, char* title)
{
    cvNamedWindow(title, 0);   //參數0表示生成的窗口大小可調整，參數1表示窗口自適應圖像而用戶不可調整，所以我選擇參數0
    cvShowImage(title, img);
    cvReleaseImage(&img);
}

3. siftfeat.c

注釋掉下面一行：

#include <unistd.h>

原因：顧名思義，unistd.h是unix std的意思，是POSIX標准定義的unix類系統定義符號常量的頭文件，所以在windows下先注釋掉。

然后注釋掉下面這兩個函數：static void arg_parse( int argc, char** argv ) 和 static void usage( char* name )

  1 static void arg_parse( int argc, char** argv )
  2 {
  3   //extract program name from command line (remove path, if present)
  4   pname = basename( argv[0] );
  5 
  6   //parse commandline options
  7   while( 1 )
  8     {
  9       char* arg_check;
 10       int arg = getopt( argc, argv, OPTIONS );
 11       if( arg == -1 )
 12     break;
 13 
 14       switch( arg )
 15     {
 16       // catch unsupplied required arguments and exit
 17     case ':':
 18       fatal_error( "-%c option requires an argument\n"        \
 19                "Try '%s -h' for help.", optopt, pname );
 20       break;
 21 
 22       // read out_file_name
 23     case 'o':
 24       if( ! optarg )
 25         fatal_error( "error parsing arguments at -%c\n"    \
 26              "Try '%s -h' for help.", arg, pname );
 27       out_file_name = optarg;
 28       break;
 29 
 30       // read out_img_name
 31     case 'm':
 32       if( ! optarg )
 33         fatal_error( "error parsing arguments at -%c\n"    \
 34              "Try '%s -h' for help.", arg, pname );
 35       out_img_name = optarg;
 36       break;
 37       
 38       // read intervals
 39     case 'i':
 40       // ensure argument provided
 41       if( ! optarg )
 42         fatal_error( "error parsing arguments at -%c\n"    \
 43              "Try '%s -h' for help.", arg, pname );
 44       
 45       // parse argument and ensure it is an integer
 46       intvls = strtol( optarg, &arg_check, 10 );
 47       if( arg_check == optarg  ||  *arg_check != '\0' )
 48         fatal_error( "-%c option requires an integer argument\n"    \
 49              "Try '%s -h' for help.", arg, pname );
 50       break;
 51       
 52       // read sigma
 53     case 's' :
 54       // ensure argument provided
 55       if( ! optarg )
 56         fatal_error( "error parsing arguments at -%c\n"    \
 57              "Try '%s -h' for help.", arg, pname );
 58       
 59       // parse argument and ensure it is a floating point number
 60       sigma = strtod( optarg, &arg_check );
 61       if( arg_check == optarg  ||  *arg_check != '\0' )
 62         fatal_error( "-%c option requires a floating point argument\n" \
 63              "Try '%s -h' for help.", arg, pname );
 64       break;
 65       
 66       // read contrast_thresh
 67     case 'c' :
 68       // ensure argument provided
 69       if( ! optarg )
 70         fatal_error( "error parsing arguments at -%c\n"    \
 71              "Try '%s -h' for help.", arg, pname );
 72       
 73       // parse argument and ensure it is a floating point number
 74       contr_thr = strtod( optarg, &arg_check );
 75       if( arg_check == optarg  ||  *arg_check != '\0' )
 76         fatal_error( "-%c option requires a floating point argument\n" \
 77              "Try '%s -h' for help.", arg, pname );
 78       break;
 79       
 80       // read curvature_thresh
 81     case 'r' :
 82       // ensure argument provided
 83       if( ! optarg )
 84         fatal_error( "error parsing arguments at -%c\n"    \
 85              "Try '%s -h' for help.", arg, pname );
 86       
 87       // parse argument and ensure it is a floating point number
 88       curv_thr = strtol( optarg, &arg_check, 10 );
 89       if( arg_check == optarg  ||  *arg_check != '\0' )
 90         fatal_error( "-%c option requires an integer argument\n"    \
 91              "Try '%s -h' for help.", arg, pname );
 92       break;
 93       
 94       // read descr_width
 95     case 'n' :
 96       // ensure argument provided
 97       if( ! optarg )
 98         fatal_error( "error parsing arguments at -%c\n"    \
 99              "Try '%s -h' for help.", arg, pname );
100       
101       // parse argument and ensure it is a floating point number
102       descr_width = strtol( optarg, &arg_check, 10 );
103       if( arg_check == optarg  ||  *arg_check != '\0' )
104         fatal_error( "-%c option requires an integer argument\n"    \
105              "Try '%s -h' for help.", arg, pname );
106       break;
107       
108       // read descr_histo_bins
109     case 'b' :
110       // ensure argument provided
111       if( ! optarg )
112         fatal_error( "error parsing arguments at -%c\n"    \
113              "Try '%s -h' for help.", arg, pname );
114       
115       // parse argument and ensure it is a floating point number
116       descr_hist_bins = strtol( optarg, &arg_check, 10 );
117       if( arg_check == optarg  ||  *arg_check != '\0' )
118         fatal_error( "-%c option requires an integer argument\n"    \
119              "Try '%s -h' for help.", arg, pname );
120       break;
121       
122       // read double_image
123     case 'd' :
124       img_dbl = ( img_dbl == 1 )? 0 : 1;
125       break;
126 
127       // read display
128     case 'x' :
129       display = 0;
130       break;
131 
132       // user asked for help
133     case 'h':
134       usage( pname );
135       exit(0);
136       break;
137 
138       // catch invalid arguments
139     default:
140       fatal_error( "-%c: invalid option.\nTry '%s -h' for help.",
141                optopt, pname );
142     }
143     }
144 
145   // make sure an input file is specified
146   if( argc - optind < 1 )
147     fatal_error( "no input file specified.\nTry '%s -h' for help.", pname );
148 
149   // make sure there aren't too many arguments
150   if( argc - optind > 1 )
151     fatal_error( "too many arguments.\nTry '%s -h' for help.", pname );
152 
153   // copy image file name from command line argument
154   img_file_name = argv[optind];
155 }

View Code

 1 static void usage( char* name )
 2 {
 3   fprintf(stderr, "%s: detect SIFT keypoints in an image\n\n", name);
 4   fprintf(stderr, "Usage: %s [options] <img_file>\n", name);
 5   fprintf(stderr, "Options:\n");
 6   fprintf(stderr, "  -h               Display this message and exit\n");
 7   fprintf(stderr, "  -o <out_file>    Output keypoints to text file\n");
 8   fprintf(stderr, "  -m <out_img>     Output keypoint image file (format" \
 9       " determined by extension)\n");
10   fprintf(stderr, "  -i <intervals>   Set number of sampled intervals per" \
11       " octave in scale space\n");
12   fprintf(stderr, "                   pyramid (default %d)\n",
13       SIFT_INTVLS);
14   fprintf(stderr, "  -s <sigma>       Set sigma for initial gaussian"    \
15       " smoothing at each octave\n");
16   fprintf(stderr, "                   (default %06.4f)\n", SIFT_SIGMA);
17   fprintf(stderr, "  -c <thresh>      Set threshold on keypoint contrast" \
18       " |D(x)| based on [0,1]\n");
19   fprintf(stderr, "                   pixel values (default %06.4f)\n",
20       SIFT_CONTR_THR);
21   fprintf(stderr, "  -r <thresh>      Set threshold on keypoint ratio of" \
22       " principle curvatures\n");
23   fprintf(stderr, "                   (default %d)\n", SIFT_CURV_THR);
24   fprintf(stderr, "  -n <width>       Set width of descriptor histogram" \
25       " array (default %d)\n", SIFT_DESCR_WIDTH);
26   fprintf(stderr, "  -b <bins>        Set number of bins per histogram" \
27       " in descriptor array\n");
28   fprintf(stderr, "                   (default %d)\n", SIFT_DESCR_HIST_BINS);
29   fprintf(stderr, "  -d               Toggle image doubling (default %s)\n",
30       SIFT_IMG_DBL == 0 ? "off" : "on");
31   fprintf(stderr, "  -x               Turn off keypoint display\n");
32 }

View Code

注釋原因：此函數是一些控制台操作，其中有許多Unix標准庫提供的函數，而我們不需要。

注釋掉main函數中的：

arg_parse( argc, argv );

注釋原因：調用了上面那個arg_parse函數。

4. imgfeatures.c

在首行加入宏定義：

#define M_PI 3.14159265358979323846

原因：報錯“M_PI”未聲明。

暫時這么多，可能還有些細節問題，但是很容易解決。

二. 運行程序

好了到目前為止可以進行試驗了，因為上文提到有4個main函數，所以可以分別運行來看看他們到底實現了什么功能。

1. siftfeat.c

利用siftfeat.c進行特征點的檢測，返回檢測到的特征點數目並在圖中標記出。

配置好的siftfeat.c文件：

 1 /*
 2   This program detects image features using SIFT keypoints. For more info,
 3   refer to:
 4   
 5   Lowe, D. Distinctive image features from scale-invariant keypoints.
 6   International Journal of Computer Vision, 60, 2 (2004), pp.91--110.
 7   
 8   Copyright (C) 2006-2012  Rob Hess <rob@iqengines.com>
 9 
10   Note: The SIFT algorithm is patented in the United States and cannot be
11   used in commercial products without a license from the University of
12   British Columbia.  For more information, refer to the file LICENSE.ubc
13   that accompanied this distribution.
14 
15   Version: 1.1.2-20100521
16 */
17 
18 #include "sift.h"
19 #include "imgfeatures.h"
20 #include "utils.h"
21 
22 #include <opencv/highgui.h>
23 
24 //#include <unistd.h> //unix 標准頭文件
25 
26 #define OPTIONS ":o:m:i:s:c:r:n:b:dxh"
27 
28 /*************************** Function Prototypes *****************************/
29 
30 static void usage( char* );
31 static void arg_parse( int, char** );
32 
33 /******************************** Globals ************************************/
34 
35 char* pname;
36 char* img_file_name = "G:\\360downloads\\pc.jpg";    //待檢測圖片的絕對路徑
37 char* out_file_name = NULL;   //導出特征點到此文件中
38 char* out_img_name = NULL;   //導出圖片的文件名
39 int intvls = SIFT_INTVLS;
40 double sigma = SIFT_SIGMA;
41 double contr_thr = SIFT_CONTR_THR;
42 int curv_thr = SIFT_CURV_THR;
43 int img_dbl = SIFT_IMG_DBL;
44 int descr_width = SIFT_DESCR_WIDTH;
45 int descr_hist_bins = SIFT_DESCR_HIST_BINS;
46 int display = 1;
47 
48 
49 /********************************** Main *************************************/
50 
51 int main( int argc, char** argv )
52 {
53   IplImage* img;
54   struct feature* features;
55   int n = 0;
56 
57   //arg_parse( argc, argv );
58 
59   fprintf( stderr, "Finding SIFT features...\n" );
60   img = cvLoadImage( img_file_name, 1 );
61   if( ! img )
62     fatal_error( "unable to load image from %s", img_file_name );
63   n = _sift_features( img, &features, intvls, sigma, contr_thr, curv_thr,
64               img_dbl, descr_width, descr_hist_bins );
65   fprintf( stderr, "Found %d features.\n", n );
66   
67   if( display )
68     {
69       draw_features( img, features, n );
70       display_big_img( img, img_file_name );
71       cvWaitKey( 0 );
72     }
73 
74   if( out_file_name != NULL )
75     export_features( out_file_name, features, n );
76 
77   if( out_img_name != NULL )
78     cvSaveImage( out_img_name, img, NULL );
79   return 0;
80 }

View Code

直接運行得到結果：

輸入圖像輸出圖像

可以看到找到了3001個特征點。

2. match.c

注意這次調用match中的main函數，所以暫時將siftfeat.c移除。

配置好的match.c文件：

  1 /*
  2   Detects SIFT features in two images and finds matches between them.
  3 
  4   Copyright (C) 2006-2012  Rob Hess <rob@iqengines.com>
  5 
  6   @version 1.1.2-20100521
  7 */
  8 
  9 #include "sift.h"
 10 #include "imgfeatures.h"
 11 #include "kdtree.h"
 12 #include "utils.h"
 13 #include "xform.h"
 14 
 15 #include <opencv/cv.h>
 16 #include <opencv/cxcore.h>
 17 #include <opencv/highgui.h>
 18 
 19 #include <stdio.h>
 20 
 21 
 22 /* the maximum number of keypoint NN candidates to check during BBF search */
 23 #define KDTREE_BBF_MAX_NN_CHKS 200
 24 
 25 /* threshold on squared ratio of distances between NN and 2nd NN */
 26 #define NN_SQ_DIST_RATIO_THR 0.49
 27 
 28 
 29 int main( int argc, char** argv )
 30 {
 31   IplImage* img1, * img2, * stacked;
 32   struct feature* feat1, * feat2, * feat;
 33   struct feature** nbrs;
 34   struct kd_node* kd_root;
 35   CvPoint pt1, pt2;
 36   double d0, d1;
 37   int n1, n2, k, i, m = 0;
 38 
 39   if( argc != 3 )
 40     fatal_error( "usage: %s <img1> <img2>", argv[0] );
 41   
 42   img1 = cvLoadImage( argv[1], 1 );
 43   if( ! img1 )
 44     fatal_error( "unable to load image from %s", argv[1] );
 45   img2 = cvLoadImage( argv[2], 1 );
 46   if( ! img2 )
 47     fatal_error( "unable to load image from %s", argv[2] );
 48   stacked = stack_imgs( img1, img2 );
 49 
 50   fprintf( stderr, "Finding features in %s...\n", argv[1] );
 51   n1 = sift_features( img1, &feat1 );
 52   fprintf( stderr, "Finding features in %s...\n", argv[2] );
 53   n2 = sift_features( img2, &feat2 );
 54   fprintf( stderr, "Building kd tree...\n" );
 55   kd_root = kdtree_build( feat2, n2 );
 56   for( i = 0; i < n1; i++ )
 57     {
 58       feat = feat1 + i;
 59       k = kdtree_bbf_knn( kd_root, feat, 2, &nbrs, KDTREE_BBF_MAX_NN_CHKS );
 60       if( k == 2 )
 61     {
 62       d0 = descr_dist_sq( feat, nbrs[0] );
 63       d1 = descr_dist_sq( feat, nbrs[1] );
 64       if( d0 < d1 * NN_SQ_DIST_RATIO_THR )
 65         {
 66           pt1 = cvPoint( cvRound( feat->x ), cvRound( feat->y ) );
 67           pt2 = cvPoint( cvRound( nbrs[0]->x ), cvRound( nbrs[0]->y ) );
 68           pt2.y += img1->height;
 69           cvLine( stacked, pt1, pt2, CV_RGB(255,0,255), 1, 8, 0 );
 70           m++;
 71           feat1[i].fwd_match = nbrs[0];
 72         }
 73     }
 74       free( nbrs );
 75     }
 76 
 77   fprintf( stderr, "Found %d total matches\n", m );
 78   display_big_img( stacked, "Matches" );
 79   cvWaitKey( 0 );
 80 
 81   /* 
 82      UNCOMMENT BELOW TO SEE HOW RANSAC FUNCTION WORKS
 83      
 84      Note that this line above:
 85      
 86      feat1[i].fwd_match = nbrs[0];
 87      
 88      is important for the RANSAC function to work.
 89   */
 90   /*
 91   {
 92     CvMat* H;
 93     IplImage* xformed;
 94     H = ransac_xform( feat1, n1, FEATURE_FWD_MATCH, lsq_homog, 4, 0.01,
 95               homog_xfer_err, 3.0, NULL, NULL );
 96     if( H )
 97       {
 98     xformed = cvCreateImage( cvGetSize( img2 ), IPL_DEPTH_8U, 3 );
 99     cvWarpPerspective( img1, xformed, H, 
100                CV_INTER_LINEAR + CV_WARP_FILL_OUTLIERS,
101                cvScalarAll( 0 ) );
102     cvNamedWindow( "Xformed", 1 );
103     cvShowImage( "Xformed", xformed );
104     cvWaitKey( 0 );
105     cvReleaseImage( &xformed );
106     cvReleaseMat( &H );
107       }
108   }
109   */
110 
111   cvReleaseImage( &stacked );
112   cvReleaseImage( &img1 );
113   cvReleaseImage( &img2 );
114   kdtree_release( kd_root );
115   free( feat1 );
116   free( feat2 );
117   return 0;
118 }

match.c

注意運行此程序的方法，以及圖片的存放路徑，具體步驟見博客。

例如我的兩張圖片為：pc1.jpg 和 pc2.jpg,此外我都是在debug形式下調試，所以不要搞錯為release。

pc1.jpg pc2.jpg

直接運行結果：

可以看到總共找出1018對匹配點對。

此外其實注意到在match.c中的main函數有一部分代碼被注釋掉了，而這段代碼調用了xform.c,即（RANSAC算法(RANdom SAmple Consensus 隨機抽樣一致)）的結果，所以將這部分代碼取消注釋后，直接執行：

/*
  Detects SIFT features in two images and finds matches between them.

  Copyright (C) 2006-2012  Rob Hess <rob@iqengines.com>

  @version 1.1.2-20100521
*/

#include "sift.h"
#include "imgfeatures.h"
#include "kdtree.h"
#include "utils.h"
#include "xform.h"

#include <opencv/cv.h>
#include <opencv/cxcore.h>
#include <opencv/highgui.h>

#include <stdio.h>


/* the maximum number of keypoint NN candidates to check during BBF search */
#define KDTREE_BBF_MAX_NN_CHKS 200

/* threshold on squared ratio of distances between NN and 2nd NN */
#define NN_SQ_DIST_RATIO_THR 0.49


int main( int argc, char** argv )
{
  IplImage* img1, * img2, * stacked;
  struct feature* feat1, * feat2, * feat;
  struct feature** nbrs;
  struct kd_node* kd_root;
  CvPoint pt1, pt2;
  double d0, d1;
  int n1, n2, k, i, m = 0;

  if( argc != 3 )
    fatal_error( "usage: %s <img1> <img2>", argv[0] );
  
  img1 = cvLoadImage( argv[1], 1 );
  if( ! img1 )
    fatal_error( "unable to load image from %s", argv[1] );
  img2 = cvLoadImage( argv[2], 1 );
  if( ! img2 )
    fatal_error( "unable to load image from %s", argv[2] );
  stacked = stack_imgs( img1, img2 );

  fprintf( stderr, "Finding features in %s...\n", argv[1] );
  n1 = sift_features( img1, &feat1 );
  fprintf( stderr, "Finding features in %s...\n", argv[2] );
  n2 = sift_features( img2, &feat2 );
  fprintf( stderr, "Building kd tree...\n" );
  kd_root = kdtree_build( feat2, n2 );
  for( i = 0; i < n1; i++ )
    {
      feat = feat1 + i;
      k = kdtree_bbf_knn( kd_root, feat, 2, &nbrs, KDTREE_BBF_MAX_NN_CHKS );
      if( k == 2 )
    {
      d0 = descr_dist_sq( feat, nbrs[0] );
      d1 = descr_dist_sq( feat, nbrs[1] );
      if( d0 < d1 * NN_SQ_DIST_RATIO_THR )
        {
          pt1 = cvPoint( cvRound( feat->x ), cvRound( feat->y ) );
          pt2 = cvPoint( cvRound( nbrs[0]->x ), cvRound( nbrs[0]->y ) );
          pt2.y += img1->height;
          cvLine( stacked, pt1, pt2, CV_RGB(255,0,255), 1, 8, 0 );
          m++;
          feat1[i].fwd_match = nbrs[0];
        }
    }
      free( nbrs );
    }

  fprintf( stderr, "Found %d total matches\n", m );
  display_big_img( stacked, "Matches" );
  cvWaitKey( 0 );

  /* 
     UNCOMMENT BELOW TO SEE HOW RANSAC FUNCTION WORKS
     
     Note that this line above:
     
     feat1[i].fwd_match = nbrs[0];
     
     is important for the RANSAC function to work.
  */
  
  {
    CvMat* H;
    IplImage* xformed;
    H = ransac_xform( feat1, n1, FEATURE_FWD_MATCH, lsq_homog, 4, 0.01,
              homog_xfer_err, 3.0, NULL, NULL );
    if( H )
      {
    xformed = cvCreateImage( cvGetSize( img2 ), IPL_DEPTH_8U, 3 );
    cvWarpPerspective( img1, xformed, H, 
               CV_INTER_LINEAR + CV_WARP_FILL_OUTLIERS,
               cvScalarAll( 0 ) );
    cvNamedWindow( "Xformed", 1 );
    cvShowImage( "Xformed", xformed );
    cvWaitKey( 0 );
    cvReleaseImage( &xformed );
    cvReleaseMat( &H );
      }
  }
  

  cvReleaseImage( &stacked );
  cvReleaseImage( &img1 );
  cvReleaseImage( &img2 );
  kdtree_release( kd_root );
  free( feat1 );
  free( feat2 );
  return 0;
}