-
Notifications
You must be signed in to change notification settings - Fork 0
/
panorama.py
331 lines (279 loc) · 11.5 KB
/
panorama.py
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
import cv2
import numpy as np
from glob import glob
import networkx as nx
import tkinter as tk
from tkinter import ttk
def sift_matching_with_homography(img1, img2):
gray1 = cv2.cvtColor(img1, cv2.COLOR_RGB2GRAY)
gray2 = cv2.cvtColor(img2, cv2.COLOR_RGB2GRAY)
max_size = 1000
H_scale = np.eye(3)
if gray1.shape[0] > max_size or gray1.shape[1] > max_size:
scale = max_size / max(gray1.shape[0], gray1.shape[1])
new_shape = (int(gray1.shape[1] * scale), int(gray1.shape[0] * scale))
H_scale = np.diag([scale, scale, 1])
gray1 = cv2.warpPerspective(gray1, H_scale, new_shape)
gray2 = cv2.warpPerspective(gray2, H_scale, new_shape)
sift = cv2.SIFT_create()
keypoints1, descriptors1 = sift.detectAndCompute(gray1, None)
keypoints2, descriptors2 = sift.detectAndCompute(gray2, None)
FLANN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50)
flann = cv2.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(descriptors1, descriptors2, k=2)
good_matches = [m for m, n in matches if m.distance < 0.7 * n.distance]
if len(good_matches) > 10:
src_pts = np.float32([keypoints1[m.queryIdx].pt for m in good_matches]).reshape(-1, 2)
dst_pts = np.float32([keypoints2[m.trainIdx].pt for m in good_matches]).reshape(-1, 2)
H, mask = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, 5.0)
if H is None or np.linalg.det(H) == 0 or np.sum(mask) < 20 or np.mean(mask) < 0.1:
return None
return np.linalg.inv(H_scale) @ H @ H_scale
else:
return None
sift = None
def get_sift_features(image):
global sift
if sift is None:
sift = cv2.SIFT_create(nfeatures=1000)
return sift.detectAndCompute(image, None)
def match_sift_features(feat1, feat2, use_knn=False):
keypoints1, descriptors1 = feat1
keypoints2, descriptors2 = feat2
if use_knn:
FLANN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50)
flann = cv2.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(descriptors1, descriptors2, k=2)
good_matches = [m for m, n in matches if m.distance < 0.7 * n.distance]
else:
bf = cv2.BFMatcher()
matches1 = bf.knnMatch(descriptors1, descriptors2, 2)
matches2 = bf.knnMatch(descriptors2, descriptors1, 2)
# Apply ratio test and cross-checking
good_matches = []
for m, n in matches1:
if m.distance < 0.75 * n.distance:
for n_match in matches2[m.trainIdx]:
if n_match.trainIdx == m.queryIdx:
good_matches.append(m)
if len(good_matches) > 10:
src_pts = np.float32([keypoints1[m.queryIdx].pt for m in good_matches]).reshape(-1, 2)
dst_pts = np.float32([keypoints2[m.trainIdx].pt for m in good_matches]).reshape(-1, 2)
H, mask = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, 5.0)
return H, np.mean(mask), np.sum(mask)
else:
return np.eye(3), 0, 0
def lightest_path(adj_matrix, source, target):
# Create a graph from the adjacency matrix
G = nx.Graph()
num_nodes = len(adj_matrix)
for i in range(num_nodes):
for j in range(i + 1, num_nodes):
if adj_matrix[i][j] != 0:
G.add_edge(i, j, weight=adj_matrix[i][j])
# Find the lightest path
lightest_path = nx.shortest_path(G, source=source, target=target, weight='weight')
# Calculate the total weight of the lightest path
total_weight = nx.shortest_path_length(G, source=source, target=target, weight='weight')
return lightest_path, total_weight
# Function to find the corners of an image
def get_image_corners(image):
h, w = image.shape[:2]
return np.array([
[0, 0, 1],
[w, 0, 1],
[w, h, 1],
[0, h, 1]
]).T
# Function to transform the corners with a given homography
def transform_corners(corners, H):
transformed_corners = H @ corners
transformed_corners /= transformed_corners[2] # Normalize by the last row
return transformed_corners[:2]
def update_progress_bar(progress_bar, value):
if progress_bar is None:
return
progress_bar['value'] = value
progress_bar.update()
def create_cache(images, with_gui=True, step_penalty = 0.1, matching_dist=10):
if with_gui:
root = tk.Tk()
root.title("Cache Initialization Progress")
step_label = tk.Label(root, text="Initializing...")
step_label.pack(padx=10,pady=10)
current_progress = ttk.Progressbar(root, orient="horizontal", length=400, mode="determinate")
current_progress.pack(padx=10,pady=10)
total_progress = ttk.Progressbar(root, orient="horizontal", length=400, mode="determinate")
total_progress.pack(padx=10,pady=10)
else:
current_progress = None
total_progress = None
N = len(images)
def update_step_label(text):
if with_gui:
step_label.config(text=text)
step_label.update()
time_ratio = 0.8
first_steps = N
second_steps = 0
for i in range(N):
for j in range(i):
second_steps += (abs(i-j) < matching_dist)
total_steps = time_ratio*first_steps + (1-time_ratio)*second_steps
# Step 1: Extracting features
features = []
update_step_label("Extracting features")
for i, image in enumerate(images):
features.append(get_sift_features(image))
cur_ratio = i / N
global_ratio = time_ratio*i/total_steps
update_progress_bar(total_progress, global_ratio * 100)
update_progress_bar(current_progress, cur_ratio * 100)
inliers = np.zeros([N, N])
ratios = np.zeros([N, N])
# 2d array of None homographies
matrices = [[None for _ in range(N)] for _ in range(N)]
# Step 2: Matching features
update_step_label("Matching features")
step = 0
for i in range(N):
for j in range(i):
if abs(i-j) > matching_dist:
continue
H, score, inlier = match_sift_features(features[i], features[j])
ratios[i, j] = score
inliers[i, j] = inlier
matrices[i][j] = H
matrices[j][i] = np.linalg.inv(H)
step += 1
cur_ratio = step / second_steps
global_ratio = (time_ratio*first_steps + (1-time_ratio)*step)/total_steps
update_progress_bar(current_progress, cur_ratio * 100)
update_progress_bar(total_progress, global_ratio * 100)
metric = np.minimum(ratios, inliers / 100)
np.seterr(divide = 'ignore')
scores = -np.log(metric + metric.T) + step_penalty
np.seterr(divide = 'warn')
cache = dict(N=N, scores=scores, matrices=matrices)
if with_gui:
root.destroy()
return cache
def stitch_images(images, cache, main_image=0, weight_threshold=1.0, target_size = [1920, 1080], with_gui=True):
if with_gui:
root = tk.Tk()
root.title("Panorama Stitching Progress")
step_label = tk.Label(root, text="Initializing...")
step_label.pack(padx=10,pady=10)
current_progress = ttk.Progressbar(root, orient="horizontal", length=400, mode="determinate")
current_progress.pack(padx=10,pady=10)
total_progress = ttk.Progressbar(root, orient="horizontal", length=400, mode="determinate")
total_progress.pack(padx=10,pady=10)
else:
current_progress = None
total_progress = None
N = len(images)
total_steps = 2
def update_step_label(text):
if with_gui:
step_label.config(text=text)
step_label.update()
N2 = cache['N']
if N != N2:
raise ValueError(f"Number of images in cache ({N2}) does not match number of images provided ({N})")
scores = cache['scores']
matrices = cache['matrices']
# Step 1: Computing homographies
update_step_label("Computing homographies")
homographies = []
min_x, min_y = np.inf, np.inf
max_x, max_y = -np.inf, -np.inf
for i in range(N):
src = i
H = np.eye(3)
if src != main_image:
path, weight = lightest_path(scores, main_image, src)
if weight > weight_threshold:
H = None
else:
for j in range(1, len(path)):
H_rel = matrices[path[j]][path[j - 1]]
H = H @ H_rel
homographies.append(H)
if H is not None:
corners = get_image_corners(images[i])
transformed_corners = transform_corners(corners, H)
min_x = min(min_x, np.min(transformed_corners[0]))
max_x = max(max_x, np.max(transformed_corners[0]))
min_y = min(min_y, np.min(transformed_corners[1]))
max_y = max(max_y, np.max(transformed_corners[1]))
cur_ratio = i / N
global_ratio = (2+cur_ratio)/total_steps
update_progress_bar(total_progress, global_ratio * 100)
update_progress_bar(current_progress, cur_ratio * 100)
offset_x = -min_x
offset_y = -min_y
H_translation = np.array([
[1, 0, offset_x],
[0, 1, offset_y],
[0, 0, 1]
])
scale_x = target_size[0] / (max_x - min_x)
scale_y = target_size[1] / (max_y - min_y)
scale = min(scale_x, scale_y)
H_scale = np.array([
[scale, 0, 0],
[0, scale, 0],
[0, 0, 1]
])
panorama_width = int(np.ceil(max_x - min_x) * scale)
panorama_height = int(np.ceil(max_y - min_y) * scale)
final_transforms = []
aligned_images = []
# Step 2: Warping images
scales = []
update_step_label("Warping images")
for i in range(N):
if homographies[i] is None:
final_transforms.append(None)
continue
scale = 1/np.linalg.norm(homographies[i][:2, :2])
scales.append(scale)
H_final = H_scale @ H_translation @ homographies[i]
im = cv2.warpPerspective(images[i], H_final, (panorama_width, panorama_height))
if len(im.shape) == 2:
im = cv2.cvtColor(im, cv2.COLOR_GRAY2RGB)
final_transforms.append(H_final)
aligned_images.append(im)
cur_ratio = i / N
global_ratio = (0+cur_ratio)/total_steps
update_progress_bar(current_progress, cur_ratio * 100)
update_progress_bar(total_progress, global_ratio * 100)
sum_image = aligned_images[0].copy()
# Step 3: Stitching images
update_step_label("Stitching images")
order = np.argsort(scales)
for i, ind in enumerate(order):
image = aligned_images[ind]
mask = ((image > 0) * 255).astype(np.uint8)
mask = cv2.erode(mask.astype(np.uint8), np.ones((5, 5), np.uint8))[:, :, 0]
sum_image[mask > 0] = image[mask > 0]
cur_ratio = i / len(aligned_images)
global_ratio = (1+cur_ratio)/total_steps
update_progress_bar(current_progress, cur_ratio * 100)
update_progress_bar(total_progress, global_ratio * 100)
if with_gui:
root.destroy()
return sum_image, final_transforms
# Example usage:
if __name__ == "__main__":
image_files = sorted(glob('output/simulated2/*'))
images = [cv2.imread(image_file) for image_file in image_files]
cache = create_cache(images)
panorama, transforms = stitch_images(images, cache)
import matplotlib.pyplot as plt
plt.imshow(cv2.cvtColor(panorama, cv2.COLOR_BGR2RGB))
plt.show()