如何使用OpenCV进行图像拼（Image Stitching）来创建全景图？

Question

我有多张重叠的图像，需要将它们拼接在一起以创建一个全景图。请编写一个Python程序使用OpenCV完成这一任务，并解释其中涉及的关键步骤和难点。特别注意如何处理图像之间的对齐、特征检测、匹配和拼接的平滑处理。

Answer 1

import cv2
import numpy as np

def stitch_images(image_paths):
    # Load images
    images = [cv2.imread(path) for path in image_paths]
    
    # Check if images are loaded successfully
    if any(image is None for image in images):
        print("Error loading images.")
        return None
    
    # Initialize SIFT detector
    sift = cv2.SIFT_create()

    # Detect keypoints and descriptors for each image
    keypoints_descriptors = [sift.detectAndCompute(image, None) for image in images]

    # Initialize BFMatcher
    bf = cv2.BFMatcher()

    # Find matches between images
    matches = [bf.knnMatch(keypoints_descriptors[i][1], keypoints_descriptors[i+1][1], k=2) 
               for i in range(len(images) - 1)]

    # Apply ratio test and extract good matches
    good_matches = []
    for m in matches:
        good = []
        for m, n in m:
            if m.distance < 0.75 * n.distance:
                good.append(m)
        good_matches.append(good)

    # Check if enough matches are found
    if any(len(g) < 10 for g in good_matches):
        print("Not enough matches found for stitching.")
        return None

    # Find homography and stitch images
    result = images[0]
    for i in range(len(images) - 1):
        src_pts = np.float32([keypoints_descriptors[i][0][m.queryIdx].pt for m in good_matches[i]]).reshape(-1, 1, 2)
        dst_pts = np.float32([keypoints_descriptors[i+1][0][m.trainIdx].pt for m in good_matches[i]]).reshape(-1, 1, 2)

        # Compute homography
        H, _ = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, 5.0)

        # Warp image
        h, w = images[i+1].shape[:2]
        result = cv2.warpPerspective(result, H, (result.shape[1] + w, result.shape[0]))
        result[0:h, 0:w] = images[i+1]

    # Show result
    cv2.imshow('Stitched Image', result)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

    return result

# Example usage
image_paths = ['image1.jpg', 'image2.jpg', 'image3.jpg']
stitch_images(image_paths)