import cv2 import numpy as np def _order_points(pts): rect = np.zeros((4, 2), dtype='float32') s = pts.sum(axis=1) rect[0] = pts[np.argmin(s)] rect[2] = pts[np.argmax(s)] d = np.diff(pts, axis=1) rect[1] = pts[np.argmin(d)] rect[3] = pts[np.argmax(d)] return rect def _apply_perspective(img, pts): rect = _order_points(pts.astype('float32')) width_top = np.linalg.norm(rect[1] - rect[0]) width_bot = np.linalg.norm(rect[2] - rect[3]) height_left = np.linalg.norm(rect[3] - rect[0]) height_right = np.linalg.norm(rect[2] - rect[1]) max_width = int(max(width_top, width_bot)) max_height = int(max(height_left, height_right)) scale = 800 / max_width target_w = 800 target_h = max(600, min(int(max_height * scale), 1400)) dst = np.array([[0, 0], [target_w, 0], [target_w, target_h], [0, target_h]], dtype='float32') M = cv2.getPerspectiveTransform(rect, dst) return cv2.warpPerspective(img, M, (target_w, target_h)) def _method1_paper_edges(img): gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) blurred = cv2.GaussianBlur(gray, (7, 7), 0) edges = cv2.Canny(blurred, 30, 100) edges = cv2.dilate(edges, None, iterations=2) contours, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) if not contours: return None img_area = img.shape[0] * img.shape[1] for cnt in sorted(contours, key=cv2.contourArea, reverse=True)[:5]: if cv2.contourArea(cnt) < img_area * 0.2: continue epsilon = 0.02 * cv2.arcLength(cnt, True) approx = cv2.approxPolyDP(cnt, epsilon, True) if len(approx) == 4: return _apply_perspective(img, approx.reshape(4, 2)) return None def _method2_grid_lines(img): gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) blurred = cv2.GaussianBlur(gray, (5, 5), 0) binary = cv2.adaptiveThreshold(blurred, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, 15, 10) h, w = binary.shape h_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (w // 8, 1)) h_lines = cv2.morphologyEx(binary, cv2.MORPH_OPEN, h_kernel, iterations=2) v_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (1, h // 8)) v_lines = cv2.morphologyEx(binary, cv2.MORPH_OPEN, v_kernel, iterations=2) grid = cv2.dilate(cv2.add(h_lines, v_lines), None, iterations=3) contours, _ = cv2.findContours(grid, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) if not contours: return None largest = max(contours, key=cv2.contourArea) if cv2.contourArea(largest) < (h * w) * 0.15: return None for source in [largest, cv2.convexHull(largest)]: epsilon = 0.02 * cv2.arcLength(source, True) approx = cv2.approxPolyDP(source, epsilon, True) if len(approx) == 4: return _apply_perspective(img, approx.reshape(4, 2)) return None def _method3_fallback(img): h, w = img.shape[:2] scale = 800 / w return cv2.resize(img, (800, int(h * scale))) def rectify_scoresheet(img: np.ndarray) -> tuple[np.ndarray, str]: """Rectifica la perspectiva de una planilla. Devuelve (imagen_rectificada, metodo_usado).""" result = _method1_paper_edges(img) if result is not None: return result, "metodo1_bordes_papel" result = _method2_grid_lines(img) if result is not None: return result, "metodo2_grilla" return _method3_fallback(img), "metodo3_fallback"