import cv2 import numpy as np def enhance_scanned_image(input_path, output_path="scanned_output.png"): img = cv2.imread(input_path) orig = img.copy() gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) blur = cv2.GaussianBlur(gray, (5, 5), 0) edged = cv2.Canny(blur, 75, 200) # Find contours contours, _ = cv2.findContours(edged, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE) contours = sorted(contours, key=cv2.contourArea, reverse=True) for c in contours: peri = cv2.arcLength(c, True) approx = cv2.approxPolyDP(c, 0.02 * peri, True) if len(approx) == 4: doc_cnt = approx break else: print("⚠️ Document edges not found, using original image.") doc_cnt = np.array([[[0,0]], [[img.shape[1],0]], [[img.shape[1], img.shape[0]]], [[0, img.shape[0]]]]) # Reorder points and apply perspective transform def reorder(pts): pts = pts.reshape(4, 2) new_pts = np.zeros((4, 2), dtype="float32") s = pts.sum(axis=1) diff = np.diff(pts, axis=1) new_pts[0] = pts[np.argmin(s)] new_pts[2] = pts[np.argmax(s)] new_pts[1] = pts[np.argmin(diff)] new_pts[3] = pts[np.argmax(diff)] return new_pts pts1 = reorder(doc_cnt) (tl, tr, br, bl) = pts1 width = max(np.linalg.norm(br - bl), np.linalg.norm(tr - tl)) height = max(np.linalg.norm(tr - br), np.linalg.norm(tl - bl)) pts2 = np.array([[0,0], [width-1,0], [width-1,height-1], [0,height-1]], dtype="float32") matrix = cv2.getPerspectiveTransform(pts1, pts2) scanned = cv2.warpPerspective(orig, matrix, (int(width), int(height))) # Convert to grayscale and apply adaptive threshold for whitening scanned_gray = cv2.cvtColor(scanned, cv2.COLOR_BGR2GRAY) scanned_clean = cv2.adaptiveThreshold(scanned_gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 10) cv2.imwrite(output_path, scanned_clean) print(f"✅ Scanned image saved as {output_path}") return output_path