准备工作
先准备两张待处理的图像,要求有重叠区域,且能检测到匹配点。
我选择了匹配比较难的两张原图:
导入相关依赖:
import cv2
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np
读入两张图片并灰度化:
img1 = mpimg.imread(r'D:/apap/temple/2.jpg')
img2 = mpimg.imread(r'D:/apap/temple/3.jpg')
#复制一些原图的副本,画图用
img1_list = []
img2_list = []
for i in range(15):
img1_list.append(img1.copy())
img2_list.append(img2.copy())
img11 = img1.copy()
img22 = img2.copy()
img1_g = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
img2_g = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
figure, ax = plt.subplots(1, 2, figsize=(16, 8))
ax[0].imshow(img1_g, cmap='gray')
ax[1].imshow(img2_g, cmap='gray')
plt.show()
用mpimg读取图片的原因是用matplotlib显示图片,如果用imread读取还存在RGB转换的问题。同时得到一些原图的副本,画特征点的函数都是改变原图的,所以为了画图显示方便多复制一些。特征检测和描述都是在灰度图上进行的,这样可以更好的排除颜色噪声。
灰度图显示结果:
SIFT
SIFT(Scale Invariant Feature Transform):尺度不变特征转换。计算量大,速度慢,在空间和尺度定位精准。所以一般应用在需要准确匹配和不考虑计算速度的情况下使用。也是图像拼接领域常用的特征检测算法。
在Python+OpenCV中,集成的SIFT可以直接获得特征点和描述子。
#可以传参限定点的个数上限
sift = cv2.xfeatures2d.SIFT_create()
keypoints_1, descriptors_1 = sift.detectAndCompute(img1_g,None)
keypoints_2, descriptors_2 = sift.detectAndCompute(img2_g,None)
绘制检测结果:
red = (255,0,0)
#color = (51,163,236)
cv2.drawKeypoints(image = img11,
outImage = img11,
keypoints = keypoints_1,
color = red)
cv2.drawKeypoints(image = img22,
outImage = img22,
keypoints = keypoints_2,
color = red)
figure, ax = plt.subplots(1, 2, figsize=(20, 20),dpi=160)
ax[0].imshow(img11)
ax[1].imshow(img22)
#第二个参数表示去掉多余空白位置
#plt.savefig("E:/Keypoint/SIFT_Extract.png", bbox_inches='tight')
plt.show()
用FLANN匹配器获得匹配点并筛选及可视化(匹配器用法链接cv2.FlannBasedMatcher):
FLANN_INDEX_KDTREE = 1 # kd树
index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50) # or pass empty dictionary
flann = cv2.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(descriptors_1, descriptors_2, k=2)
good_matches = []
for m,n in matches:
if m.distance < 0.75*n.distance:
good_matches.append([m])
img3 = cv2.drawMatchesKnn(img11,keypoints_1,img22,keypoints_2,good_matches,None,flags