From 460d007dc5d637a62d99dc40c6e9640a8a9aa799 Mon Sep 17 00:00:00 2001
From: milesagraham <miles.graham98@gmail.com>
Date: Mon, 20 May 2024 09:45:35 +0100
Subject: [PATCH] added cuboid (#15)

---
 src/ttmask/__init__.py |  1 +
 src/ttmask/cuboid.py   | 50 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 51 insertions(+)
 create mode 100644 src/ttmask/cuboid.py

diff --git a/src/ttmask/__init__.py b/src/ttmask/__init__.py
index 1c00a37..e3e1304 100644
--- a/src/ttmask/__init__.py
+++ b/src/ttmask/__init__.py
@@ -12,4 +12,5 @@
 from ._cli import cli
 from .sphere import sphere
 from .cylinder import cylinder
+from .cuboid import cuboid
 
diff --git a/src/ttmask/cuboid.py b/src/ttmask/cuboid.py
new file mode 100644
index 0000000..e1f8a2e
--- /dev/null
+++ b/src/ttmask/cuboid.py
@@ -0,0 +1,50 @@
+import numpy as np
+import einops
+import typer
+from typing import Tuple
+from ._cli import cli
+from typing_extensions import Annotated
+from scipy.ndimage import distance_transform_edt
+import mrcfile
+
+
+@cli.command(name='cuboid')
+def cuboid(
+    boxsize: int = typer.Option(...),
+    cuboid_sidelengths: Annotated[Tuple[float, float, float], typer.Option()] = (None, None, None),
+    soft_edge_size: float = typer.Option(...),
+):
+    c = boxsize // 2
+    center = np.array([c, c, c])
+    mask = np.zeros(shape=(boxsize, boxsize, boxsize), dtype=np.float32)
+
+    # 3d positions of all voxels
+    positions = np.indices([boxsize, boxsize, boxsize])
+    positions = einops.rearrange(positions, 'zyx d h w -> d h w zyx')
+
+    # calculate the distance between the center and every pixel position
+    print(center.shape)
+    print(positions.shape)
+
+    print('calculating distance')
+    difference = np.abs(positions - center)  # (100, 100, 100, 3)
+    # z = difference[:, :, :, 0]
+    # y = difference[:, :, :, 1]
+    # x = difference[:, :, :, 2]
+    # idx_z = z < cuboid_sidelengths[0] / 2
+    # idx_y = y < cuboid_sidelengths[1] / 2
+    # idx_x = x < cuboid_sidelengths[2] / 2
+
+    # mask[np.logical_not(idx)] = 1 #if you wanted to do opposite for whatever reason
+
+    idx = np.all(difference < (np.array(cuboid_sidelengths) / 2), axis=-1)
+
+    mask[idx] = 1
+
+    distance_from_edge = distance_transform_edt(mask == 0)
+    boundary_pixels = (distance_from_edge <= soft_edge_size) & (distance_from_edge != 0)
+    normalised_distance_from_edge = (distance_from_edge[boundary_pixels] / soft_edge_size) * np.pi
+
+    mask[boundary_pixels] = (0.5 * np.cos(normalised_distance_from_edge) + 0.5)
+
+    mrcfile.write("cuboid.mrc", mask, voxel_size=4, overwrite=True)