課程作業的一個題目,找了代碼加了註釋。
import numpy as np
import cv2
class Stitcher:
def stitch(self, images, ratio=0.75, reprojThresh=4.0,
showMatches=False):
# 檢測出關鍵點,局部不變描述符
(imageB, imageA) = images
(kpsA, featuresA) = self.detectAndDescribe(imageA)
(kpsB, featuresB) = self.detectAndDescribe(imageB)
print("關鍵點個數",len(kpsA),len(kpsB))
# 特徵匹配
M = self.matchKeypoints(kpsA, kpsB,
featuresA, featuresB, ratio, reprojThresh)
# 如果特徵匹配返回None
if M is None:
return None
# 將圖像粘合在一起
(matches, H, status) = M
# 根據單應性矩陣進行矯正圖片
result = cv2.warpPerspective(imageA, H,
(imageA.shape[1] + imageB.shape[1], imageA.shape[0]))
# imageA.shape[1]=400,imageB.shape[1]=400,imageA.shape[0]=533
# result.shape[0]=533,result.shape[1]=800
result[0:imageB.shape[0], 0:imageB.shape[1]] = imageB
# 顯示匹配線
if showMatches:
vis = self.drawMatches(imageA, imageB, kpsA, kpsB, matches,
status)
return (result, vis)
# 單獨返回一個
return result
#接收照片,檢測關鍵點和提取局部不變特徵
#用到了高斯差分(Difference of Gaussian (DoG))關鍵點檢測,和SIFT特徵提取
#detectAndCompute方法用來處理提取關鍵點和特徵
#返回一系列的關鍵點
def detectAndDescribe(self, image):
# 將圖片轉化爲灰度圖像
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# 提取特徵點
descriptor = cv2.xfeatures2d.SIFT_create()
(kps, features) = descriptor.detectAndCompute(image, None)
# print(kps) # 關鍵點
print(features.shape[0],features.shape[1]) # 長度爲128維的特徵向量
# 將關鍵點的座標pt存入numpy
kps = np.float32([kp.pt for kp in kps])
return (kps, features)
#matchKeypoints方法需要四個參數,第一張圖片的關鍵點和特徵向量,第二張圖片的關鍵點特徵向量。
#David Lowe’s ratio測試變量和RANSAC重投影門限也應該被提供。
def matchKeypoints(self, kpsA, kpsB, featuresA, featuresB,
ratio, reprojThresh):
matcher = cv2.DescriptorMatcher_create("BruteForce")
rawMatches = matcher.knnMatch(featuresA, featuresB, 2) # 最近鄰算法設置K=2
matches = []
# for m in rawMatches:
# print(m[0].distance,m[1].distance)
print("------------------------------")
# 循環遍歷匹配點
for m in rawMatches:
# Lowe’s ratio測試,用來確定高質量的特徵匹配
if len(m) == 2 and m[0].distance < m[1].distance * ratio:
# 將第一張圖像的下標值和第二張圖像的下標值存入
matches.append((m[0].trainIdx, m[0].queryIdx))
# print(matches)
# print(len(matches))
# 將標註位置存入numpy
if len(matches) > 4:
# construct the two sets of points
ptsA = np.float32([kpsA[i] for (_, i) in matches])
ptsB = np.float32([kpsB[i] for (i, _) in matches])
# 計算單應性矩陣
# 其中H爲求得的單應性矩陣矩陣
# status則返回一個列表來表徵匹配成功的特徵點。
(H, status) = cv2.findHomography(ptsA, ptsB, cv2.RANSAC,
reprojThresh)
return (matches, H, status)
return None
#連線畫出兩幅圖的匹配
def drawMatches(self, imageA, imageB, kpsA, kpsB, matches, status):
# initialize the output visualization image
(hA, wA) = imageA.shape[:2]
(hB, wB) = imageB.shape[:2]
# 三通道照片
vis = np.zeros((max(hA, hB), wA + wB, 3), dtype="uint8")
vis[0:hA, 0:wA] = imageA
vis[0:hB, wA:] = imageB
for ((trainIdx, queryIdx), s) in zip(matches, status):
if s == 1:
ptA = (int(kpsA[queryIdx][0]), int(kpsA[queryIdx][1]))
ptB = (int(kpsB[trainIdx][0]) + wA, int(kpsB[trainIdx][1]))
cv2.line(vis, ptA, ptB, (0, 255, 0), 1)
return vis
if __name__ == '__main__':
# 加載圖片
imageA = cv2.imread('./hw2/building_02.jpg')
imageB = cv2.imread('./hw2/building_03.jpg')
# 調整圖片寬度
# imageA = imutils.resize(imageA, width=400)
# imageB = imutils.resize(imageB, width=400)
# showMatches=True 展示兩幅圖像特徵的匹配,返回vis
stitcher = Stitcher()
(result, vis) = stitcher.stitch([imageA, imageB], showMatches=True)
# vis = imutils.resize(imageA, width=800,height=800)
# result = imutils.resize(imageB, width=800,height=800)
cv2.imwrite('./vis1.jpg', vis)
cv2.imwrite('./result.jpg', result)
參考:
https://www.cnblogs.com/lqerio/p/11601951.html
https://blog.csdn.net/weixin_44072651/article/details/89262277
https://blog.csdn.net/xull88619814/article/details/81587595