-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathadaptive_weigthing_quantization.m
153 lines (121 loc) · 6.38 KB
/
adaptive_weigthing_quantization.m
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
function [ histogram ] = adaptive_weigthing_quantization(local_feature,edge_feature,visual_vocabulary,window_size,paramater_orientation,paramater_distance )
%ADAPTIVE_WEIGTHING_QUANTIZATION 此处显示有关此函数的摘要
% 此处显示详细说明
% local_feature:65536*36
% edge_feature:256*256
% visual_vocabulary:2000*36
% histogram:1*65536(每一个值表示词典中的某一个单词)
%% **********************************************************************************************************************
% 计算local feature时只计算了edge feature中的兴趣点（值为非零的点）
% 同理，量化local feature时也只需要量化计算了local feature的兴趣点即可
%% **********************************************************************************************************************
%PDIST2 Pairwise distance between two sets of observations.
% D = PDIST2(X,Y) returns a matrix D containing the Euclidean distances
% between each pair of observations in the MX-by-N data matrix X and
% MY-by-N data matrix Y. Rows of X and Y correspond to observations,
% and columns correspond to variables. D is an MX-by-MY matrix, with the
% (I,J) entry equal to distance between observation I in X and
% observation J in Y.
%% 为了计算方便，调整 edge_feature:256*256——> 1*65536
% A =
%
% 1 2 3 4
% 5 6 7 8
% 9 10 11 12
% 13 14 15 16
%
% reshape(A',1,[])
% ans =
%
% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
%% 计算梯度值
[~,~,~,gradient_orientation] = calculate_gradient(edge_feature); %返回的是一个256*256矩阵，矩阵中的元素的值是梯度的方向
edge_map = edge_feature; % 一个256*256的矩阵
codebook = visual_vocabulary; %一个N*36维的矩阵
local_feature_index_matrix = zeros(1,size(local_feature,2));
for k = 1:size(local_feature,2)
local_feature_index_matrix(k)=k;
end
local_feature_index_matrix = reshape(local_feature_index_matrix',[256 256]);
%%
%% 记录量化后的索引：256*256的矩阵
histogram = zeros(size(edge_feature,1), size(edge_feature,2));
%%
for i=1:size(edge_map,1)
for j=1:size(edge_map,2)
% fprintf('%d \n', i);
if edge_map(i,j)~=0 %即当前点为兴趣点,计算当前点的local feature应该被量化为哪个codeword
%% 计算窗口大小
row_start = i-window_size; %起始行
if(row_start<1)
row_start=1;
end
row_end = i+window_size; %终止行
if(row_end>size(gradient_orientation,1))
row_end=size(gradient_orientation,1);
end
column_start = j-window_size;%起始列
if(column_start<1)
column_start=1;
end
column_end = j+window_size;%终止列
if(column_end>size(gradient_orientation,2))
column_end=size(gradient_orientation,2);
end
%% 获得梯度方向矩阵：一个窗口大小
gradient_orientation_matrix = gradient_orientation(row_start:row_end,column_start:column_end);
%% 计算梯度方向距离矩阵：一个窗口大小
% 在新的窗口中，当前兴趣点的坐标为：（i-row_start+1,j-colume_start+1）
row_index = i-row_start+1;
column_index = j-column_start+1;
gradient_orientation_distance_matrix =zeros(size(gradient_orientation,1),size(gradient_orientation,2));
for m=1:size(gradient_orientation,1)
for n=1:size(gradient_orientation,2)
%计算梯度方向差值:方向插值的平方（用弧度计算）
gradient_orientation_distance_matrix(m,n)= ((gradient_orientation_matrix(m,n) - gradient_orientation_matrix(row_index,column_index))/(2*pi))^2;
end
end
%% 计算兴趣点距离矩阵：一个窗口大小
distance_matrix = zeros(size(gradient_orientation,1),size(gradient_orientation,2));
for m=1:size(gradient_orientation,1)
for n=1:size(gradient_orientation,2)
%计算距离插值：坐标索引差值的平方
distance_matrix(m,n)=(m-row_index)^2 +(n-column_index)^2;
end
end
%% 计算权重矩阵：一个窗口大小
weighting_matrix = zeros(size(gradient_orientation,1),size(gradient_orientation,2));
for m=1:size(gradient_orientation,1)
for n=1:size(gradient_orientation,2)
weighting_matrix(m,n)=( exp(-gradient_orientation_distance_matrix(m,n)/(paramater_orientation^2)) * exp(-distance_matrix(m,n)/(paramater_distance^2)))/N;
%计算距离插值:paramater_orientation和paramater_distance是公式中的两个参数，N是一个正则项
end
end
%% 提取边缘矩阵:一个窗口大小
edge_map_matrix_window = edge_map(row_start:row_end,column_start:column_end);
%% 提取局部特征索引矩阵：一个窗口大小
local_feature_index_matrix_window = local_feature_index_matrix(row_start:row_end,column_start:column_end);
%% 量化特征
codeword_index = quantize_local_feature(weigthing_matrix,local_feature_index_matrix_window,edge_map_matrix_window);
%% 记录codeword的索引号
histogram(i,j) = codeword_index; % 记录当前兴趣点量化后的索引号
else
histogram(i,j) = 0; %非兴趣点不需要被量化，直接标识为0
end
end
end
histogram = reshape(histogram,[],1)';
function [I] = quantize_local_feature(weighting_matrix,local_feature_index_matrix_window,edge_map_matrix_window)
%获得窗口内的兴趣点的local feature:(window_size*window_size)*36
local_feature_matrix_window = zeros(size(edge_map_matrix_window,1)*size(edge_map_matrix_window,2),36);
index_matrix_window = reshape(local_feature_index_matrix_window',1,[]);
for r=1:size(index_matrix_window,2)
local_feature_matrix_window(r)=local_feature(index_matrix_window(r));
end
weighting_matrix = reshape(weighting_matrix',1,[]);
E = pdist2(local_feature_matrix_window,visual_vocabulary);
local_feature_codeword_distance_matrix = repmat(weighting_matrix,1,size(index_matrix_window,2)).*E;
final_distance_matrix = sum(local_feature_codeword_distance_matrix,1); %得到local feature到codeword的最后的距离值
[~,I] = min(final_distance_matrix);
end
end