Python实现对多张图片进行全景图像拼接

import cv2 import numpy as np import os import re from tqdm import tqdm # 用于显示进度条 def resize_image(img, max_dimension=1024): """ 将图片缩小到最大边长为 max_dimension """ height, width = img.shape[:2] if max(height, width) > max_dimension: scale = max_dimension / max(height, width) img = cv2.resize(img, (int(width * scale), int(height * scale)), interpolation=cv2.INTER_AREA) return img def find_homography_and_warp(img1, img2): """ 计算单应性矩阵并变换第二张图片 """ # 使用 AKAZE 特征检测算法 akaze = cv2.AKAZE_create() kp1, desc1 = akaze.detectAndCompute(img1, None) kp2, desc2 = akaze.detectAndCompute(img2, None) # 使用 FLANN 匹配器 FLANN_INDEX_LSH = 6 index_params = dict(algorithm=FLANN_INDEX_LSH, table_number=6, key_size=12, multi_probe_level=1) search_params = dict(checks=50) flann = cv2.FlannBasedMatcher(index_params, search_params) # 匹配描述符 matches = flann.knnMatch(desc1, desc2, k=2) # 筛选好的匹配 good = [] for m, n in matches: if m.distance < 0.7 * n.distance: good.append(m) # 至少需要 4 个匹配点 if len(good) < 4: raise ValueError("匹配点不足，无法计算单应性矩阵") # 获取匹配点的位置 pt1 = np.float32([kp1[m.queryIdx].pt for m in good]).reshape(-1, 1, 2) pt2 = np.float32([kp2[m.trainIdx].pt for m in good]).reshape(-1, 1, 2) # 计算单应性矩阵 H, mask = cv2.findHomography(pt2, pt1, cv2.RANSAC, ransacReprojThreshold=4.0) # 使用单应性矩阵变换第二张图片 height = img1.shape[0] + img2.shape[0] width = img1.shape[1] + img2.shape[1] canvas = np.zeros((height, width, 3), np.uint8) warped_img2 = cv2.warpPerspective(img2, H, (canvas.shape[1], canvas.shape[0])) return warped_img2 def multi_band_blending(img1, img2): """ 使用多频段融合消除鬼影 """ # 创建多频段融合器 blender = cv2.detail_MultiBandBlender() blender.prepare((0, 0, max(img1.shape[1], img2.shape[1]), max(img1.shape[0], img2.shape[0]))) # 创建单通道的 mask mask1 = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY) mask1 = (mask1 > 0).astype(np.uint8) * 255 # 将非零区域设置为 255 mask2 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY) mask2 = (mask2 > 0).astype(np.uint8) * 255 # 将非零区域设置为 255 # 将图像添加到融合器 blender.feed(img1.astype(np.uint8), mask1, (0, 0)) # 使用 uint8 类型和单通道 mask blender.feed(img2.astype(np.uint8), mask2, (0, 0)) # 使用 uint8 类型和单通道 mask # 融合图像 result, _ = blender.blend(None, None) return result.astype(np.uint8) def crop_black_borders(img): """ 裁剪黑色边缘 """ gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) _, thresh = cv2.threshold(gray, 1, 255, cv2.THRESH_BINARY) contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) if not contours: return img x, y, w, h = cv2.boundingRect(contours[0]) return img[y:y+h, x:x+w] def stitch_images(image_list): """ 拼接多张图片 """ if len(image_list) < 2: raise ValueError("至少需要两张图片进行拼接") # 初始化拼接结果为第一张图片 result = image_list[0] # 依次拼接剩余的图片 for i in tqdm(range(1, len(image_list)), desc="拼接进度"): img2 = image_list[i] try: warped_img2 = find_homography_and_warp(result, img2) result = multi_band_blending(result, warped_img2) # 使用多频段融合 except ValueError as e: print(f"跳过第 {i+1} 张图片：{e}") continue # 裁剪黑色边缘 result = crop_black_borders(result) return result def load_images_from_folder(folder, max_dimension=1024): """ 从文件夹中加载所有图片，并按数字编号排序 """ images = [] filenames = [] # 获取文件夹中所有图片文件名 for filename in os.listdir(folder): if filename.endswith(('.jpg', '.jpeg', '.png', '.bmp')): filenames.append(filename) # 按数字编号排序 filenames.sort(key=lambda f: int(re.sub('\D', '', f))) # 加载图片并缩小 for filename in filenames: img_path = os.path.join(folder, filename) img = cv2.imread(img_path) if img is not None: img = resize_image(img, max_dimension) # 缩小图片 images.append(img) print(f"已加载图片: {filename}") else: print(f"无法加载图片: {filename}") return images # 从文件夹中加载图片 folder_path = "images" # 替换为你的图片文件夹路径 images = load_images_from_folder(folder_path, max_dimension=1024) # 限制图片最大边长为 1024 if len(images) < 2: print("文件夹中至少需要两张图片进行拼接") else: # 拼接图片 final_image = stitch_images(images) # 显示拼接结果 cv2.imshow('Stitched Image', final_image) # 保存拼接结果 output_path = "hecheng.jpg" cv2.imwrite(output_path, final_image) print(f"拼接结果已保存为: {output_path}") cv2.waitKey(0) cv2.destroyAllWindows()