init

.gitignore

@@ -0,0 +1,3 @@
+.idea/
+*.npy
+fig.png

reconstruct.py

@@ -0,0 +1,27 @@
+import math
+
+import numpy as np
+from matplotlib import pyplot as plt
+
+width = 100
+
+grid = np.zeros((width, width), dtype=np.uint64)
+
+all_data: dict = np.load("out.npy", allow_pickle=True).item()
+print(all_data)
+for y in range(width):
+    print(y)
+    for x in range(width):
+        for start, bins in all_data.items():
+            pixpos = np.array([x, y])
+            myradians = math.atan2(start[0] - x, start[1] - y)
+            angle = np.degrees(myradians)
+            if not np.isnan(angle):
+                bin = int(round(angle) % 360)
+                grid[x, y] += bins[bin]
+        # exit()
+grid = grid.T
+# grid[grid > 200] = 200
+plt.imshow(grid)
+plt.colorbar()
+plt.show()

simulate.py

@@ -0,0 +1,52 @@
+import math
+
+import imageio
+import matplotlib.pyplot as plt
+import numpy as np
+from scipy.ndimage import zoom, gaussian_filter
+
+img = imageio.imread('input_slight_blur.png')
+print(img)
+# img = zoom(img, 3)
+width = 100
+numstarts = 6
+i = np.linspace(0, 360, numstarts, endpoint=False)
+i += np.random.randint(-20, 20, numstarts)
+r = width / 2.3
+ang = np.radians(i)
+xstart = np.cos(ang) * r * np.random.randint(80, 110, numstarts) / 100 + width / 2
+ystart = np.sin(ang) * r * np.random.randint(80, 110, numstarts) / 100 + width / 2
+print(xstart, ystart)
+plt.scatter(xstart, ystart, c="red")
+plt.imshow(img)
+plt.show()
+
+all_data = {}
+show = True
+for i in range(numstarts):
+    pos = (xstart[i], ystart[i])
+    start = np.array(pos)
+
+    bins = np.zeros(360)
+    for y, row in enumerate(img):
+        for x, value in enumerate(row):
+            pixpos = np.array([x, y])
+            myradians = math.atan2(start[0] - x, start[1] - y)
+            angle = np.degrees(myradians)
+            # print(angle)
+            # exit()
+            if not np.isnan(angle):
+                bin = int(round(angle) % 360)
+                bins[bin] += value
+            else:
+                print(pixpos)
+    bins = gaussian_filter(bins, sigma=1)
+    all_data[pos] = bins
+    print(pos)
+    if show:
+        plt.plot(np.arange(0, 360), bins)
+        plt.xlabel("angle")
+        plt.savefig("fig.png")
+        plt.show()
+        show = False
+np.save("out", all_data)