Newer
Older
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
import time
import cv2
import numpy as np
import heapq
import matplotlib.pyplot as plt
from scipy.ndimage import convolve
from skimage.filters import gaussian
from skimage.feature import canny
#### Helper functions ####
def neighbors_8(x, y, width, height):
"""Return the 8-connected neighbors of (x, y)."""
for nx in (x-1, x, x+1):
for ny in (y-1, y, y+1):
if 0 <= nx < width and 0 <= ny < height:
if not (nx == x and ny == y):
yield nx, ny
def dijkstra(cost_img, seed):
"""
Dijkstra's algorithm on a 2D grid, using cost_img as the per-pixel cost.
Args:
cost_img (np.array): 2D array of costs (float).
seed (tuple): (x, y) starting coordinate.
Returns:
dist (np.float32): array of minimal cumulative cost from seed to each pixel.
parent (np.int32): array storing predecessor of each pixel for path reconstruction.
"""
height, width = cost_img.shape
# Initialize dist and parent
dist = np.full((height, width), np.inf, dtype=np.float32)
dist[seed[1], seed[0]] = 0.0
parent = -1 * np.ones((height, width, 2), dtype=np.int32)
visited = np.zeros((height, width), dtype=bool)
pq = [(0.0, seed[0], seed[1])] # (distance, x, y)
while pq:
curr_dist, cx, cy = heapq.heappop(pq)
if visited[cy, cx]:
continue
visited[cy, cx] = True
for nx, ny in neighbors_8(cx, cy, width, height):
if visited[ny, nx]:
continue
# We can take an average or sum—here, let's just sum the cost
move_cost = 0.5 * (cost_img[cy, cx] + cost_img[ny, nx])
ndist = curr_dist + move_cost
if ndist < dist[ny, nx]:
dist[ny, nx] = ndist
parent[ny, nx] = (cx, cy)
heapq.heappush(pq, (ndist, nx, ny))
return dist, parent
def backtrack_path(parent, start, end):
"""
Reconstruct path from 'end' back to 'start' using 'parent' array.
Args:
parent (np.array): shape (H, W, 2), storing (px, py) for each pixel.
start (tuple): (x, y) start coordinate.
end (tuple): (x, y) end coordinate.
Returns:
path (list of (x, y)): from start to end inclusive.
"""
path = []
current = end
while True:
path.append(current)
if current == start:
break
px, py = parent[current[1], current[0]]
current = (px, py)
path.reverse()
return path
def compute_cost(image, sigma=3.0, epsilon=1e-5):
smoothed_img = gaussian(image, sigma=sigma)
canny_img = canny(smoothed_img)
cost_img = 1 / (canny_img + epsilon)
return cost_img, canny_img
def load_image(path, type):
# Load image
if type == 'gray':
img = cv2.imread(path, cv2.IMREAD_GRAYSCALE)
if img is None:
raise FileNotFoundError(f"Could not read {path}")
elif type == 'color':
img = cv2.imread(path, cv2.IMREAD_COLOR)
if img is None:
raise FileNotFoundError(f"Could not read {path}")
else:
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
else:
raise ValueError("type must be 'gray' or 'color'")
return img
def downscale(img, points, scale_percent):
if scale_percent == 100:
return img, (tuple(points[0]), tuple(points[1]))
else:
width = int(img.shape[1] * scale_percent / 100)
height = int(img.shape[0] * scale_percent / 100)
new_dimensions = (width, height)
# Downsample the image
downsampled_img = cv2.resize(img, new_dimensions, interpolation=cv2.INTER_AREA)
### SCALE POINTS
# Original image dimensions
original_width = img.shape[1]
original_height = img.shape[0]
# Downsampled image dimensions
downsampled_width = width
downsampled_height = height
# Scaling factors
scale_x = downsampled_width / original_width
scale_y = downsampled_height / original_height
# Original points
seed_xy = tuple(points[0])
target_xy = tuple(points[1])
# Scale the points
scaled_seed_xy = (int(seed_xy[0] * scale_x), int(seed_xy[1] * scale_y))
scaled_target_xy = (int(target_xy[0] * scale_x), int(target_xy[1] * scale_y))
return downsampled_img, (scaled_seed_xy, scaled_target_xy)
# Define the following
image_path = './tests/slice_60_volQ.png'
image_type = 'gray' # 'gray' or 'color'
downscale_factor = 100 # % of original size wanted
points_path = './tests/LiveWireEndPoints.npy'
# Load image
image = load_image(image_path, image_type)
# Load points
points = np.int0(np.round(np.load(points_path)))
# Downscale image and points
scaled_image, scaled_points = downscale(image, points, downscale_factor)
seed, target = scaled_points
# Compute cost image
cost_image, canny_img = compute_cost(scaled_image)
# Find path and time it
start_time = time.time()
dist, parent = dijkstra(cost_image, seed)
path = backtrack_path(parent, seed, target)
end_time = time.time()
print(f"Elapsed time for pathfinding: {end_time - start_time:.3f} seconds")
color_img = cv2.cvtColor(scaled_image, cv2.COLOR_GRAY2BGR)
for (x, y) in path:
color_img[y, x] = (0, 0, 255) # red (color of path)
plt.figure(figsize=(20,8))
plt.subplot(1,2,1)
plt.title("Cost Image")
plt.imshow(canny_img, cmap='gray')
plt.subplot(1,2,2)
plt.title("Path from Seed to Target")
plt.imshow(color_img[..., ::-1]) # BGR->RGB for plotting
plt.show()