-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathsaliency.java
357 lines (342 loc) · 19.5 KB
/
saliency.java
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
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
/*
* To change this license header, choose License Headers in Project Properties.
* To change this template file, choose Tools | Templates
* and open the template in the editor.
*/
package saliency;
import static java.lang.Double.max;
import static java.lang.Math.abs;
import org.opencv.core.Core;
import static org.opencv.core.CvType.CV_16UC1;
import static org.opencv.core.CvType.CV_8UC1;
import static org.opencv.core.CvType.CV_64F;
import static org.opencv.core.CvType.CV_64FC1;
import org.opencv.core.Mat;
import org.opencv.core.Size;
import static org.opencv.highgui.Highgui.imread;
import static org.opencv.highgui.Highgui.imwrite;
import org.opencv.imgproc.Imgproc;
/**
*
* @author YuanXiquan
*/
public class saliency {
/**
* @param args the command line arguments
*/
public static void main(String[] args) {
// TODO code application logic here
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
Mat input_img = imread("input_img/sea.jpg");
//fot temp test start
Imgproc.resize(input_img, input_img, new Size(1980,1080), 0, 0, Imgproc.INTER_LINEAR);
//fot temp test end
if(input_img.cols() == 0){
return;
}
//benchmark
///////////////////////step 1 : Extraction of Early Visual Deatures///////////////////////////////
//intensity image: intensity_img
Mat intensity_img = new Mat(input_img.rows(), input_img.cols(), CV_16UC1);
//intensity = (R+G+B)/3
int img_width = intensity_img.cols();
int img_height = intensity_img.rows();
int x, y;
int i, c, s;
int max_intensity = 0;
for(x = 0; x < img_width; x++){
for(y = 0; y < img_height; y++){
int temp_intensity = ((int)input_img.get(y,x)[0] + (int)input_img.get(y,x)[1] + (int)input_img.get(y,x)[2])/3;
intensity_img.put(y,x, temp_intensity);
if(max_intensity < temp_intensity){
max_intensity = temp_intensity;
}
}
}
//create Guassian pyramid for intensity
Mat[] i_gaussian_pyramid = new Mat[9];
i_gaussian_pyramid[0] = intensity_img.clone();
for(i = 0; i < 8; i++){
i_gaussian_pyramid[i+1] = i_gaussian_pyramid[i].clone();
Imgproc.pyrDown(i_gaussian_pyramid[i+1], i_gaussian_pyramid[i+1], new Size());
}
//create intensity feature map using center-surround differences
Mat[][] intensity_feature_map = new Mat[3][2];
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
intensity_feature_map[c][s] = center_surround.main(i_gaussian_pyramid[c+2], i_gaussian_pyramid[s+c+5], 0);
}
}
//benchmark
//imwrite("intensity_feature_map_00.bmp", intensity_feature_map[0][0]);
//get normalized color image by I.
Mat norm_input_img = input_img.clone();
norm_input_img.convertTo(norm_input_img, CV_64F);
for(x = 0; x < img_width; x++){
for(y = 0; y < img_height; y++){
//normalization is only applied at the locations where I is larger than 1/10 of its maximum over entire image
double[] temp = new double[3];
if(intensity_img.get(y,x)[0]>(max_intensity/10)){
temp[0] = norm_input_img.get(y,x)[0]/intensity_img.get(y,x)[0];
temp[1] = norm_input_img.get(y,x)[1]/intensity_img.get(y,x)[0];
temp[2] = norm_input_img.get(y,x)[2]/intensity_img.get(y,x)[0];
norm_input_img.put(y,x,temp);
}else{
temp[0] = 0;
temp[1] = 0;
temp[2] = 0;
norm_input_img.put(y,x,temp);
}
}
}
//get R G B Y(Yellow) single color channel images
Mat r_img = new Mat(input_img.rows(), input_img.cols(), CV_64FC1);
Mat g_img = new Mat(input_img.rows(), input_img.cols(), CV_64FC1);
Mat b_img = new Mat(input_img.rows(), input_img.cols(), CV_64FC1);
Mat y_img = new Mat(input_img.rows(), input_img.cols(), CV_64FC1);
//[0]: b [1]:g [2]:r
for(x = 0; x < img_width; x++){
for(y = 0; y < img_height; y++){
//R = min(0,r-(g+b)/2)
double temp_chroma = max(0, (norm_input_img.get(y,x)[2]-(norm_input_img.get(y,x)[1]+norm_input_img.get(y,x)[0])/2));
r_img.put(y,x, temp_chroma);
//G = max(0,g-(r+b)/2)
temp_chroma = max(0, (norm_input_img.get(y,x)[1]-(norm_input_img.get(y,x)[2]+norm_input_img.get(y,x)[0])/2));
g_img.put(y,x, temp_chroma);
//B = max(0,b-(r+g)/2)
temp_chroma = max(0, (norm_input_img.get(y,x)[0]-(norm_input_img.get(y,x)[2]+norm_input_img.get(y,x)[1])/2));
b_img.put(y,x, temp_chroma);
//Y = max(0,(r+g)/2-|r-g|/2-b)
temp_chroma = max(0, ((norm_input_img.get(y,x)[2]+norm_input_img.get(y,x)[1])/2 -abs(norm_input_img.get(y,x)[2]+norm_input_img.get(y,x)[1])/2-norm_input_img.get(y,x)[0]));
y_img.put(y,x, temp_chroma);
}
}
//create Gaussian pyramid for 4 color channels
Mat[] b_gaussian_pyramid = new Mat[9];
b_gaussian_pyramid[0] = b_img.clone();
for(i = 0; i < 8; i++){
b_gaussian_pyramid[i+1] = b_gaussian_pyramid[i].clone();
Imgproc.pyrDown(b_gaussian_pyramid[i+1], b_gaussian_pyramid[i+1], new Size());
}
Mat[] g_gaussian_pyramid = new Mat[9];
g_gaussian_pyramid[0] = g_img.clone();
for(i = 0; i < 8; i++){
g_gaussian_pyramid[i+1] = g_gaussian_pyramid[i].clone();
Imgproc.pyrDown(g_gaussian_pyramid[i+1], g_gaussian_pyramid[i+1], new Size());
}
Mat[] r_gaussian_pyramid = new Mat[9];
r_gaussian_pyramid[0] = r_img.clone();
for(i = 0; i < 8; i++){
r_gaussian_pyramid[i+1] = r_gaussian_pyramid[i].clone();
Imgproc.pyrDown(r_gaussian_pyramid[i+1], r_gaussian_pyramid[i+1], new Size());
}
Mat[] y_gaussian_pyramid = new Mat[9];
y_gaussian_pyramid[0] = y_img.clone();
for(i = 0; i < 8; i++){
y_gaussian_pyramid[i+1] = y_gaussian_pyramid[i].clone();
Imgproc.pyrDown(y_gaussian_pyramid[i+1], y_gaussian_pyramid[i+1], new Size());
}
//create color feature map using center-surround differences
//RG(c,s) = |(R(c)-G(c))(-)(G(c)-R(c))|
Mat[][] rg_feature_map = new Mat[3][2];
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat r_minus_g = r_gaussian_pyramid[c+2].clone();
Core.subtract(r_gaussian_pyramid[c+2], g_gaussian_pyramid[c+2], r_minus_g);
Mat g_minus_r = g_gaussian_pyramid[s+c+5].clone();
Core.subtract(g_gaussian_pyramid[s+c+5], r_gaussian_pyramid[s+c+5], g_minus_r);
rg_feature_map[c][s] = center_surround.main(r_minus_g, g_minus_r, 1);
}
}
//BY(c,s) = |(B(c)-Y(c))(-)(Y(c)-B(c))|
Mat[][] by_feature_map = new Mat[3][2];
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat b_minus_g = b_gaussian_pyramid[c+2].clone();
Core.subtract(b_gaussian_pyramid[c+2], y_gaussian_pyramid[c+2], b_minus_g);
Mat y_minus_b = y_gaussian_pyramid[s+c+5].clone();
Core.subtract(y_gaussian_pyramid[s+c+5], b_gaussian_pyramid[s+c+5], y_minus_b);
by_feature_map[c][s] = center_surround.main(b_minus_g, y_minus_b, 1);
}
}
//benchmark
//create oriented Gabor pyramid from intensity image
int kernel_size = 10;//31;//adjust value according to reference
double sigma = 3;//default: σ = 0.56 λ. the larger σ, the support of the Gabor function and the number of visible parallel excitatory and inhibitory stripe zones increases.
double[] theta = new double[4];
theta[0] = 0;
theta[1] = Math.PI/4;
theta[2] = Math.PI/2;
theta[3] = Math.PI*3/4;
double lambda = 5;//36; minimum 3
double gamma = 0.5;//0.02;
// double psi = 0;
Mat[][] gabor_pyramid = new Mat[4][9];
int theta_index;
for(theta_index = 0; theta_index < 4; theta_index++){
Mat gabor_kernel = Imgproc.getGaborKernel(new Size(kernel_size,kernel_size), sigma, theta[theta_index], lambda, gamma);
//gabor_pyramid[theta_index][0] = intensity_img.clone();
for(i = 0; i < 9; i++){
//gabor_pyramid[theta_index][i] = gabor_pyramid[theta_index][i].clone();
gabor_pyramid[theta_index][i] = i_gaussian_pyramid[i].clone();
Imgproc.filter2D(i_gaussian_pyramid[i], gabor_pyramid[theta_index][i], -1, gabor_kernel);
//Imgproc.resize(gabor_pyramid[theta_index][i], gabor_pyramid[theta_index][i], new Size(), 0.5, 0.5, Imgproc.INTER_AREA);
}
}
//imwrite("gabor_pyramid_01.bmp", gabor_pyramid[0][1]);
//imwrite("gabor_pyramid_11.bmp", gabor_pyramid[1][1]);
//imwrite("gabor_pyramid_21.bmp", gabor_pyramid[2][1]);
//imwrite("gabor_pyramid_31.bmp", gabor_pyramid[3][1]);
//imwrite("gabor_pyramid_03.bmp", gabor_pyramid[0][3]);
//get orientation feature map using center-surround differences
Mat[][][] orientation_feature_map = new Mat[4][3][2];
for(theta_index = 0; theta_index < 4; theta_index++){
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
orientation_feature_map[theta_index][c][s] = center_surround.main(gabor_pyramid[theta_index][c+2], gabor_pyramid[theta_index][s+c+5], 0);
}
}
}
//benchmark
//imwrite("orientation_test_00.bmp", orientation_feature_map[0][0][0]);
///////////////////////step 2 : the saliency map///////////////////////////////
//get intensity conspicuity map
Mat intensity_conspicuity_map = Mat.zeros(intensity_feature_map[2][0].size(), CV_16UC1);
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat norm_out = map_norm.main(intensity_feature_map[c][s]);
Mat resized_feature_map = Mat.zeros(intensity_feature_map[2][0].size(), CV_16UC1);
Imgproc.resize(norm_out, resized_feature_map, intensity_feature_map[2][0].size(), 0, 0, Imgproc.INTER_LINEAR);
Core.addWeighted(intensity_conspicuity_map, 1, resized_feature_map, 1.0/6, 0, intensity_conspicuity_map);
/*if(c == 0 && s == 0){
imwrite("in.bmp", intensity_feature_map[c][s]);
imwrite("map_norm.bmp",norm_out);
imwrite("resized_feature_map.bmp", resized_feature_map);
}*/
}
}
//benchmark
//Core.normalize(intensity_conspicuity_map, intensity_conspicuity_map, 0, 255, Core.NORM_MINMAX);
//imwrite("intensity_conspicuity_map.bmp", intensity_conspicuity_map);
//get color conspicuity map
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Core.normalize(rg_feature_map[c][s], rg_feature_map[c][s], 0, 255, Core.NORM_MINMAX);
rg_feature_map[c][s].convertTo(rg_feature_map[c][s], CV_16UC1);
Core.normalize(by_feature_map[c][s], by_feature_map[c][s], 0, 255, Core.NORM_MINMAX);
by_feature_map[c][s].convertTo(by_feature_map[c][s], CV_16UC1);
}
}
//imwrite("test_rg.bmp",rg_feature_map[0][0]);
Mat color_conspicuity_map = Mat.zeros(rg_feature_map[2][0].size(), CV_16UC1);
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat norm_out = map_norm.main(rg_feature_map[c][s]);
Mat resized_feature_map = Mat.zeros(rg_feature_map[2][0].size(), CV_16UC1);
Imgproc.resize(norm_out, resized_feature_map, rg_feature_map[2][0].size(), 0, 0, Imgproc.INTER_LINEAR);
Core.addWeighted(color_conspicuity_map, 1, resized_feature_map, 1.0/12, 0, color_conspicuity_map);
norm_out = map_norm.main(by_feature_map[c][s]);
resized_feature_map = Mat.zeros(by_feature_map[2][0].size(), CV_16UC1);
Imgproc.resize(norm_out, resized_feature_map, by_feature_map[2][0].size(), 0, 0, Imgproc.INTER_LINEAR);
Core.addWeighted(color_conspicuity_map, 1, resized_feature_map, 1.0/12, 0, color_conspicuity_map);
}
}
//benchmark
//get orientation conspicuity map
Mat orientation_conspicuity_map_0 = Mat.zeros(orientation_feature_map[0][2][0].size(), CV_16UC1);
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat norm_out = map_norm.main(orientation_feature_map[0][c][s]);
Mat resized_feature_map = Mat.zeros(orientation_feature_map[0][2][0].size(), CV_16UC1);
Imgproc.resize(norm_out, resized_feature_map, orientation_feature_map[0][2][0].size(), 0, 0, Imgproc.INTER_LINEAR);
Core.addWeighted(orientation_conspicuity_map_0, 1, resized_feature_map, 1.0/6, 0, orientation_conspicuity_map_0);
}
}
Mat orientation_conspicuity_map_1 = Mat.zeros(orientation_feature_map[1][2][0].size(), CV_16UC1);
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat norm_out = map_norm.main(orientation_feature_map[1][c][s]);
Mat resized_feature_map = Mat.zeros(orientation_feature_map[1][2][0].size(), CV_16UC1);
Imgproc.resize(norm_out, resized_feature_map, orientation_feature_map[1][2][0].size(), 0, 0, Imgproc.INTER_LINEAR);
Core.addWeighted(orientation_conspicuity_map_1, 1, resized_feature_map, 1.0/6, 0, orientation_conspicuity_map_1);
}
}
Mat orientation_conspicuity_map_2 = Mat.zeros(orientation_feature_map[2][2][0].size(), CV_16UC1);
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat norm_out = map_norm.main(orientation_feature_map[2][c][s]);
Mat resized_feature_map = Mat.zeros(orientation_feature_map[2][2][0].size(), CV_16UC1);
Imgproc.resize(norm_out, resized_feature_map, orientation_feature_map[2][2][0].size(), 0, 0, Imgproc.INTER_LINEAR);
Core.addWeighted(orientation_conspicuity_map_2, 1, resized_feature_map, 1.0/6, 0, orientation_conspicuity_map_2);
}
}
Mat orientation_conspicuity_map_3 = Mat.zeros(orientation_feature_map[3][2][0].size(), CV_16UC1);
for(c = 0; c < 3; c++){
for(s = 0; s <2; s++){
Mat norm_out = map_norm.main(orientation_feature_map[3][c][s]);
Mat resized_feature_map = Mat.zeros(orientation_feature_map[3][2][0].size(), CV_16UC1);
Imgproc.resize(norm_out, resized_feature_map, orientation_feature_map[3][2][0].size(), 0, 0, Imgproc.INTER_LINEAR);
Core.addWeighted(orientation_conspicuity_map_3, 1, resized_feature_map, 1.0/6, 0, orientation_conspicuity_map_3);
}
}
Mat orientation_conspicuity_map = Mat.zeros(orientation_feature_map[0][2][0].size(), CV_16UC1);
Core.addWeighted(orientation_conspicuity_map, 1, map_norm.main(orientation_conspicuity_map_0), 1.0/4, 0, orientation_conspicuity_map);
Core.addWeighted(orientation_conspicuity_map, 1, map_norm.main(orientation_conspicuity_map_1), 1.0/4, 0, orientation_conspicuity_map);
Core.addWeighted(orientation_conspicuity_map, 1, map_norm.main(orientation_conspicuity_map_2), 1.0/4, 0, orientation_conspicuity_map);
Core.addWeighted(orientation_conspicuity_map, 1, map_norm.main(orientation_conspicuity_map_3), 1.0/4, 0, orientation_conspicuity_map);
//benchmark
Mat saliency = Mat.zeros(intensity_conspicuity_map.size(), CV_16UC1);
Core.addWeighted(saliency, 1, map_norm.main(intensity_conspicuity_map), 1.0/3, 0, saliency);
Core.addWeighted(saliency, 1, map_norm.main(color_conspicuity_map), 1.0/3, 0, saliency);
Core.addWeighted(saliency, 1, map_norm.main(orientation_conspicuity_map), 1.0/3, 0, saliency);
//benchmark
Core.normalize(saliency, saliency, 0, 255, Core.NORM_MINMAX);
//fot temp test
Imgproc.resize(saliency, saliency, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("saliency.bmp", saliency);
Core.normalize(intensity_conspicuity_map, intensity_conspicuity_map, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(intensity_conspicuity_map, intensity_conspicuity_map, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("intensity_conspicuity_map.bmp", intensity_conspicuity_map);
Core.normalize(color_conspicuity_map, color_conspicuity_map, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(color_conspicuity_map, color_conspicuity_map, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("color_conspicuity_map.bmp", color_conspicuity_map);
Core.normalize(orientation_conspicuity_map, orientation_conspicuity_map, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(orientation_conspicuity_map, orientation_conspicuity_map, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("orientation_conspicuity_map.bmp", orientation_conspicuity_map);
Imgproc.resize(input_img, input_img, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("input_img.bmp", input_img);
//for testing algorithm
/*
Mat temp1 = Mat.zeros(intensity_conspicuity_map.size(), CV_16UC1);
temp1 = map_norm.main(intensity_conspicuity_map);
Core.normalize(temp1, temp1, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(temp1, temp1, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("intensity.bmp", temp1);
temp1 = map_norm.main(color_conspicuity_map);
Core.normalize(temp1, temp1, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(temp1, temp1, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("color.bmp", temp1);
temp1 = map_norm.main(orientation_conspicuity_map);
Core.normalize(temp1, temp1, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(temp1, temp1, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("orientation.bmp", temp1);
Mat temp2 = Mat.zeros(orientation_conspicuity_map_0.size(), CV_16UC1);
temp2 = map_norm.main(orientation_conspicuity_map_0);
Core.normalize(temp2, temp2, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(temp2, temp2, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("orientation_conspicuity_map_0.bmp", temp2);
temp2 = map_norm.main(orientation_conspicuity_map_1);
Core.normalize(temp2, temp2, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(temp2, temp2, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("orientation_conspicuity_map_1.bmp", temp2);
temp2 = map_norm.main(orientation_conspicuity_map_2);
Core.normalize(temp2, temp2, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(temp2, temp2, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("orientation_conspicuity_map_2.bmp", temp2);
temp2 = map_norm.main(orientation_conspicuity_map_3);
Core.normalize(temp2, temp2, 0, 255, Core.NORM_MINMAX);
Imgproc.resize(temp2, temp2, new Size(720,480), 0, 0, Imgproc.INTER_LINEAR);
imwrite("orientation_conspicuity_map_3.bmp", temp2);
*/
}
}