External/SnapOtter
1
0
Fork 0
mirror of https://github.com/snapotter-hq/SnapOtter.git synced 2026-08-03 07:46:42 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix(ai): tune scratch detection thresholds and remove erosive open step

Browse Source 
- Two-gate threshold: Otsu >= 60 uses Otsu; 40-59 uses fixed 100
  (catches strong scratches on borderline images)
- Remove morphological OPEN after component filtering: it was eroding
  thin scratch lines that were correctly detected
- Lower Otsu gate from 60 to 40 to avoid false-negating borderline images
This commit is contained in:
SnapOtter
2026-05-13 17:17:32 +08:00
parent f1a4c94375
commit a03d4f34ff
1 changed files with 6 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/ai/python/restore.py
+6 -5
View File
@@ -91,16 +91,17 @@ def detect_scratches(img_bgr, _sensitivity=None):
response_u8 = np.clip(response, 0, 255).astype(np.uint8)
otsu_thresh, mask = cv2.threshold(response_u8, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
if otsu_thresh < 60:
if otsu_thresh < 40:
return np.zeros_like(gray)
# When Otsu is borderline, use a high fixed threshold to only catch strong scratches
if otsu_thresh < 60:
_, mask = cv2.threshold(response_u8, 100, 255, cv2.THRESH_BINARY)
# Connected component filtering
mask = _filter_components(mask, h * w)
# Post-processing
kernel_open = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (3, 3))
mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel_open)
# Connect nearby scratch segments (no opening - it erodes thin scratch lines)
kernel_close = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (7, 7))
mask = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel_close)