Update ocr_engine.py

ocr_engine.py

@@ -1,40 +1,41 @@
 import cv2
 import easyocr
-import re
-from dblur import Deblurrer
 import numpy as np
-from PIL import Image
+import re
 
-# Initialize EasyOCR reader and Restormer deblurrer
+# Initialize EasyOCR reader
 reader = easyocr.Reader(['en'], gpu=False)
-restormer = Deblurrer(model_name="restormer")
 
-def deblur_image(pil_img):
-    img_np = np.array(pil_img)
-    deblurred_np = restormer.deblur(img_np)
-    return Image.fromarray(deblurred_np)
+def sharpen_image(img_cv):
+    kernel = np.array([[0, -1, 0],
+                       [-1, 5,-1],
+                       [0, -1, 0]])
+    return cv2.filter2D(img_cv, -1, kernel)
 
 def extract_weight_from_image(pil_img):
-    # Deblur and convert to OpenCV format
-    pil_img = deblur_image(pil_img)
-    img_cv = cv2.cvtColor(np.array(pil_img), cv2.COLOR_RGB2GRAY)
+    # Convert PIL to OpenCV
+    img_cv = cv2.cvtColor(np.array(pil_img), cv2.COLOR_RGB2BGR)
+
+    # Resize and sharpen
+    img_cv = cv2.resize(img_cv, None, fx=2.0, fy=2.0, interpolation=cv2.INTER_CUBIC)
+    img_cv = sharpen_image(img_cv)
 
-    # Resize and apply thresholding
-    img = cv2.resize(img_cv, None, fx=2.0, fy=2.0, interpolation=cv2.INTER_CUBIC)
-    img = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
-                                cv2.THRESH_BINARY, 11, 3)
+    # Convert to grayscale
+    img_gray = cv2.cvtColor(img_cv, cv2.COLOR_BGR2GRAY)
 
+    # Apply adaptive threshold and morphology
+    img_thresh = cv2.adaptiveThreshold(img_gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
+                                       cv2.THRESH_BINARY, 11, 3)
     kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (2, 2))
-    img = cv2.morphologyEx(img, cv2.MORPH_CLOSE, kernel)
+    img_clean = cv2.morphologyEx(img_thresh, cv2.MORPH_CLOSE, kernel)
 
     # OCR
-    results = reader.readtext(img)
-
+    results = reader.readtext(img_clean)
     filtered = [text for _, text, conf in results if len(text.strip()) <= 5 and conf > 0.4]
     raw_text = " ".join(filtered)
-    matches = re.findall(r"\d+\.\d+|\d+", raw_text)
 
+    matches = re.findall(r"\d+\.\d+|\d+", raw_text)
     weight = matches[0] if matches else "0.0"
     confidence = int(results[0][2] * 100) if results else 0
 
-    return weight, confidence, raw_text, pil_img  # return deblurred image too
+    return weight, confidence, raw_text, img_cv  # Return processed image