基于OpenCV的图像拼接技术实践与示例代码解析

目录

• 背景
• 实现步骤
• 一、读取文件
• 二、单应性矩阵计算
• 三、图像拼接
• 总结

img1 = cv2.imread("map1.png")
img2 = cv2.imread("map2.png")

img1 = cv2.resize(img1, (640, 480))
img2 = cv2.resize(img2, (640, 480))

input = np.hstack((img1, img2))
cv2.imshow("input", input)
cv2.waitKey(0)

def get_homo(img1, img2):
    # 1实现
    sift = cv2.xfeatures2d.SIFT_create()
    # 2实现
    k1, p1 = sift.detectAndCompute(img1, None)
    k2, p2 = sift.detectAndCompute(img2, None)
    # 3实现
    bf = cv2.BFMatcher()
    # 4实现
    matches = bf.knnMatch(p1, p2, k=2)
    # 5实现
    good = []
    for m1, m2 in matches:
        if m1.distance <0.8 * m2.distance:
            good.append(m1)
    # 6实现
    if len(good) > 8:
        img1_pts = []
        img2_pts = []
        for m in good:
            img1_pts.append(k1[m.queryIdx].pt)
            img2_pts.append(k2[m.trainIdx].pt)
        img1_pts = np.float32(img1_pts).reshape(-1, 1, 2)
        img2_pts = np.float32(img2_pts).reshape(-1, 1, 2)
        H, mask = cv2.findHomography(img1_pts, img2_pts, cv2.RANSAC, 5.0)
        return H
    else:
        print("piints is not enough 8!")
        exit()

def stitch_img(img1, img2, H):
    # 1实现
    h1, w1 = img1.shape[:2]
    h2, w2 = img2.shape[:2]
    img1_point = np.float32([[0,0], [0,h1], [w1,h1], [w1,0]]).reshape(-1, 1, 2)
    img2_point = np.float32([[0,0], [0,h2], [w2,h2], [w2,0]]).reshape(-1, 1, 2)
    # 2实现
    img1_trans = cv2.perspectiveTransform(img1_point, H)
    # 将img1变换后的角点与img2原来的角点做拼接
    result_point = np.concatenate((img2_point, img1_trans), axis=0)
    # 获得拼接后图像x,y的最小值
    [x_min, y_min] = np.int32(result_point.min(axis=0).ravel()-0.5)
    # 获得拼接后图像x,y的最大值
    [x_max, y_max] = np.int32(result_point.max(axis=0).ravel()+0.5)
    # 平移距离
    trans_dist = [-x_min, -y_min]
    # 构建一个齐次平移矩阵
    trans_array = np.array([[1, 0, trans_dist[0]],
                            [0, 1, trans_dist[1]],
                            [0, 0, 1]])
    # 平移和单应性变换
    res_img = cv2.warpPerspective(img1, trans_array.dot(H), (x_max-x_min, y_max-y_min))
    # 3实现
    res_img[trans_dist[1]:trans_dist[1]+h2,
            trans_dist[0]:trans_dist[0]+w2] = img2
    return res_img

H = get_homo(img1, img2)
res_img = stitch_img(img1, img2, H)
# 4实现
cv2.imshow("result", res_img)
cv2.waitKey(0)