利用Opencv实现图像融合拼接--以柑橘为例

运行环境：

• python3.6
• opencv3.4.15
• Opencv-contrib-python3.4.15
  

import cv2
import numpy as np

# 图像处理展示
def cvshow(name, img):
    cv2.imshow(name, img)
    cv2.waitKey(0)
    # cv2.destroyAllWindows()

# SURF特征点检测
def surf_kp(image):
    # 灰度图
    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    # 设置Hessian矩阵值，值越大，检测出特征点越精确
    surf = cv2.xfeatures2d.SURF_create(hessianThreshold=15000)
    # 提取出surf特征点和坐标
    kp, des = surf.detectAndCompute(image, None)
    # 灰度图特征点描述
    kp_image = cv2.drawKeypoints(gray_image, kp,None, flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
    return kp_image, kp, des

# 特征点匹配
def get_good_match(des1, des2):

    # FlannBasedMatcher特征匹配算法（最近邻搜索）
    FLANN_INDEX_KDTREE = 0  # 建立FLANN匹配器的参数
    indexParams = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)  # 配置索引，密度树的数量为5
    searchParams = dict(checks=50)  # 指定递归次数
    flann = cv2.FlannBasedMatcher(indexParams, searchParams)
    matches = flann.knnMatch(des1, des2, k=2)  # des1为模板图，des2为匹配图
    matches = sorted(matches, key=lambda x: x[0].distance / x[1].distance)
    good = []
    for m, n in matches:
        # 如果第一个邻近距离比第二个邻近距离的0.75倍小，则保留
        if m.distance < 0.75 * n.distance:
            good.append(m)
    return good


# 图像拼接融合
def surfimg_rightlignment(img_right, img_left):
    _, kp1, des1 = surf_kp(img_right)
    _, kp2, des2 = surf_kp(img_left)
    goodMatch = get_good_match(des1, des2)

    # 当筛选项的匹配对大于8对时：计算视角变换矩阵
    if len(goodMatch) > 8:
        # 获取匹配对的点坐标
        ptsA = np.float32([kp1[m.queryIdx].pt for m in goodMatch]).reshape(-1, 1, 2)
        ptsB = np.float32([kp2[m.trainIdx].pt for m in goodMatch]).reshape(-1, 1, 2)

        #  该函数的作用就是先用RANSAC选择最优的四组配对点，再计算H矩阵。H为3*3矩阵
        ransacReprojThreshold = 4
        H, statuRANSACs = cv2.findHomography(ptsA, ptsB, cv2.RANSAC, ransacReprojThreshold)

        # 实现透视变换转换图像融合
        result = cv2.warpPerspective(img_right, H, (img_right.shape[1] + img_left.shape[1], img_right.shape[0]),flags=cv2.INTER_NEAREST)
        result_medium = result
        cvshow('result_medium', result_medium)

        # 将左边图像映射
        result[0:img_left.shape[0], 0:img_left.shape[1]] = img_left
      
        return result
    else:
        print("not enough matches are found")

def main(left_img,right_img):
    # 读取拼接图片（注意图片左右的放置）
    img_right = cv2.imread(right_img)
    img_left = cv2.imread(left_img)

    # 按比例缩小图片
    img_right = cv2.resize(img_right, None, fx=1, fy=1)

    # 保证两张图一样大
    img_left = cv2.resize(img_left, (img_right.shape[1], img_right.shape[0]))

    # 特征点检测
    kpimg_right, kp1, des1 = surf_kp(img_right)
    kpimg_left, kp2, des2 = surf_kp(img_left)

    # 同时显示原图和特征点检测后的图
    cvshow('img_left_kp', np.hstack((img_left, kpimg_left)))
    cvshow('img_right_kp', np.hstack((img_right, kpimg_right)))

    goodMatch = get_good_match(des2,des1)

    # 特征连线配对
    all_goodmatch_img = cv2.drawMatches(img_left,kp2,img_right, kp1,goodMatch, None, flags=2)

    # 筛选匹配度较高的匹配对
    goodmatch_img = cv2.drawMatches( img_left, kp2,img_right, kp1, goodMatch[:18], None, flags=2)

    # 展示所有匹配对
    cvshow('Keypoint_Matches_ALL', all_goodmatch_img)

    # 展示筛选后的匹配对
    cvshow('Keypoint_Matches_good', goodmatch_img)

    # 图像融合拼接
    result = surfimg_rightlignment(img_right, img_left)
    cvshow('Result', result)

if __name__ == '__main__':
    main('images/2-1-1.png','images/2-2-2.png')
    

左边图像

右边图像