-
Notifications
You must be signed in to change notification settings - Fork 0
/
SPKNNSearch.c
163 lines (144 loc) · 4.65 KB
/
SPKNNSearch.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
#include "macros.h"
#include "KDTree.h"
#include "SPBPriorityQueue.h"
#include "math.h"
/**
* This function is the recursive function for the actual KNN search.
* It searches for all the neighbors of the SPPoint feat in the KD-Tree kd,
* while maintaining and updating a Bounded Priority Queue of the size KNN.
*
* @param kdtree The KD-Tree to be searched on.
* @param point Query for the search
* @param bpq The SPBPQueue to be updated after the search of
*
* @return None
*/
void knnRecursive(SPKDTreeNode kd, SPPoint feat, SPBPQueue bpq) {
if (!kd)
return;
//From here, kd is not NULL so no need to check output of inner functions
SPListElement elem = NULL;
bool left = false, equal, lessOrEqual;
double val, p_dim;
/** If the node is a leaf - try to enqueue the element with the distance
* from feat. If it is bigger than all the points already in the bpq
* than the element is not added.
*/
if (isLeaf(kd)) {
elem = spListElementCreate(getIndex(kd), getDistFromPoint(kd, feat));
spBPQueueEnqueue(bpq, elem);
spListElementDestroy(elem); //In any case elem needs to be freed because a copy is made in enqueue
elem = NULL;
return;
}
val = getVal(kd); // median of the current node, also kd is not NULL here
p_dim = spPointGetAxisCoor(feat, getDim(kd)); // P[curr.dim]
/** check if val <= p_dim **/
equal = doubleEquals(val, p_dim);
lessOrEqual = (equal) ? equal : val < p_dim;
/***************************/
/** First, continue the recursion downward the KDTree based on val **/
if (lessOrEqual) {
knnRecursive(getLeftSubtree(kd), feat, bpq);
left = true; //to know it was left
} else
knnRecursive(getRightSubtree(kd), feat, bpq);
/** Then, check the conditions for recursion continuation
* on the second branch - if the queue isn't full, than obviously
* continue to fill it until its full.
* Or, |curr.val - P[curr.dim]| < bpq.maxPriority.
*/
if (!spBPQueueIsFull(bpq) || fabs(val - p_dim) < spBPQueueMaxValue(bpq)) {
if (left)
knnRecursive(getRightSubtree(kd), feat, bpq);
else
knnRecursive(getLeftSubtree(kd), feat, bpq);
}
}
int* findKNearestNeighbors(SPKDTreeNode kdtree, SPPoint feat, SPConfig config) {
/** Initializations **/
SP_CONFIG_MSG _conf_msg = SP_CONFIG_SUCCESS, *conf_msg = &_conf_msg;
declareLogMsg();
if (!kdtree || !feat || !config) {
InvalidError();
return NULL;
}
int i,
k = spConfigGetKNN(config, conf_msg), *ret;
returnIfConfigMsg(NULL)
SPListElement tmp = NULL;
/* Creation of the Bounded Priority Queue */
SPBPQueue bpq = spBPQueueCreate(k);
if (!bpq) {
MallocError()
return NULL;
}
/** Main call to the KNN Search **/
knnRecursive(kdtree, feat, bpq);
/*********************************/
/* returned indices array */
ret = (int*) malloc(k * sizeof(int));
if (!ret) {
MallocError()
spBPQueueDestroy(bpq);
return NULL;
}
/* Copy to the indices array */
for (i = 0; i < k; i++) {
tmp = spBPQueuePeek(bpq);
ret[i] = spListElementGetIndex(tmp);
spListElementDestroy(tmp); //gets copied for the peek
tmp = NULL;
spBPQueueDequeue(bpq);
}
spBPQueueDestroy(bpq);
return ret;
}
int* getClosestImages(SPKDTreeNode kdtree, SPConfig config, SPPoint* q_features,
int q_numOfFeats) {
declareLogMsg();
declareConfMsg();
if(!kdtree || !config || !q_features || q_numOfFeats<0) {
InvalidError()
return NULL;
}
int* knn = NULL;
int i,
j;
int knn_size = spConfigGetKNN(config, conf_msg);
int numOfImages = spConfigGetNumOfImages(config, conf_msg);
int numOfSimilarImages = spConfigGetNumOfSimilarImages(config, conf_msg);
int *img_near_cnt = (int*) calloc(numOfImages, sizeof(int));
int *similar_images = (int*) calloc(numOfSimilarImages, sizeof(int));
if (!img_near_cnt || !similar_images) {
MallocError()
return NULL;
}
//for each point in the query image, find k-nearest neighbors
printInfo(KNN_DO);
for (i = 0; i < q_numOfFeats; i++) {
knn = findKNearestNeighbors(kdtree, q_features[i], config);
if (!knn) {
printError(KNN_FAIL);
return NULL;
}
//count image indices related to neighbors just found
for (j = 0; j < knn_size; j++)
img_near_cnt[knn[j]]++;
free(knn);
knn = NULL;
}
//return the k nearest images based on img_near_cnt array
// assumes numOfSimilarImages << n
printInfo(CLOS_IMGS);
for (i = 0; i < numOfSimilarImages; i++) {
similar_images[i] = 0;
for (j = 0; j < numOfImages; j++) {
if (img_near_cnt[j] > img_near_cnt[similar_images[i]])
similar_images[i] = j;
}
img_near_cnt[similar_images[i]] = 0;
}
free(img_near_cnt);
return similar_images;
}