File size: 1,900 Bytes
1d197a4 | 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 | """Frame preprocessing utilities."""
import numpy as np
def ensure_uint8(images: np.ndarray) -> np.ndarray:
"""Convert any numeric image stack to uint8 with min-max scaling."""
if images.dtype == np.uint8:
return images
images = images.astype(np.float32)
min_val = float(images.min())
max_val = float(images.max())
if max_val <= min_val:
return np.zeros(images.shape, dtype=np.uint8)
scaled = (images - min_val) * (255.0 / (max_val - min_val))
return scaled.astype(np.uint8)
def apply_center_black_circle(
images: np.ndarray,
diameter: int = 30,
center_scale: float = 0.0,
edge_scale: float = 0.5,
) -> np.ndarray:
"""Apply a soft radial attenuation mask centered in each frame.
The attenuation is linear in radius from ``center_scale`` (at center)
to ``edge_scale`` (at circle boundary), with no attenuation outside.
"""
if images.ndim != 3:
raise ValueError(f"Expected images shape (frames, rows, cols), got {images.shape}")
radius = max(1, int(diameter // 2))
center_scale = float(np.clip(center_scale, 0.0, 1.0))
edge_scale = float(np.clip(edge_scale, 0.0, 1.0))
h, w = images.shape[1], images.shape[2]
cy, cx = h // 2, w // 2
yy, xx = np.ogrid[:h, :w]
dist = np.sqrt((yy - cy) ** 2 + (xx - cx) ** 2).astype(np.float32)
scale = np.ones((h, w), dtype=np.float32)
inside = dist <= radius
t = np.zeros((h, w), dtype=np.float32)
t[inside] = dist[inside] / float(radius)
scale[inside] = center_scale + t[inside] * (edge_scale - center_scale)
masked = images.astype(np.float32) * scale[np.newaxis, :, :]
if np.issubdtype(images.dtype, np.integer):
info = np.iinfo(images.dtype)
masked = np.clip(np.rint(masked), info.min, info.max).astype(images.dtype)
else:
masked = masked.astype(images.dtype)
return masked
|