脚本之家,脚本语言编程技术及教程分享平台!
分类导航

Python|VBS|Ruby|Lua|perl|VBA|Golang|PowerShell|Erlang|autoit|Dos|bat|

服务器之家 - 脚本之家 - Python - Python图片检索之以图搜图

Python图片检索之以图搜图

2021-11-18 09:28代码小白的成长 Python

由于很多论文里面的测试图片没有标号,就不能确定它们是Testset数据集中哪几张图片.为了能解决这个问题,需要完成以图片去搜索整个数据集文件目录的任务.本文有非常详细的图文示例,需要的朋友可以参考下

一、待搜索图

Python图片检索之以图搜图

二、测试集

Python图片检索之以图搜图

三、new_similarity_compare.py

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
# -*- encoding=utf-8 -*-
 
from image_similarity_function import *
import os
import shutil
 
# 融合相似度阈值
threshold1 = 0.70
# 最终相似度较高判断阈值
threshold2 = 0.95
 
 
# 融合函数计算图片相似度
def calc_image_similarity(img1_path, img2_path):
    """
    :param img1_path: filepath+filename
    :param img2_path: filepath+filename
    :return: 图片最终相似度
    """
 
    similary_ORB = float(ORB_img_similarity(img1_path, img2_path))
    similary_phash = float(phash_img_similarity(img1_path, img2_path))
    similary_hist = float(calc_similar_by_path(img1_path, img2_path))
    # 如果三种算法的相似度最大的那个大于0.7,则相似度取最大,否则,取最小。
    max_three_similarity = max(similary_ORB, similary_phash, similary_hist)
    min_three_similarity = min(similary_ORB, similary_phash, similary_hist)
    if max_three_similarity > threshold1:
        result = max_three_similarity
    else:
        result = min_three_similarity
 
    return round(result, 3)
 
 
if __name__ == '__main__':
 
    # 搜索文件夹
    filepath = r'D:\Dataset\cityscapes\leftImg8bit\val\frankfurt'
 
    #待查找文件夹
    searchpath = r'C:\Users\Administrator\Desktop\cityscapes_paper'
 
    # 相似图片存放路径
    newfilepath = r'C:\Users\Administrator\Desktop\result'
 
    for parent, dirnames, filenames in os.walk(searchpath):
        for srcfilename in filenames:
            img1_path = searchpath +"\\"+ srcfilename
            for parent, dirnames, filenames in os.walk(filepath):
                for i, filename in enumerate(filenames):
                    print("{}/{}: {} , {} ".format(i+1, len(filenames), srcfilename,filename))
                    img2_path = filepath + "\\" + filename
                    # 比较
                    kk = calc_image_similarity(img1_path, img2_path)
                    try:
                        if kk >= threshold2:
                            # 将两张照片同时拷贝到指定目录
                            shutil.copy(img2_path, os.path.join(newfilepath, srcfilename[:-4] + "_" + filename))
                    except Exception as e:
                        # print(e)
                        pass

四、image_similarity_function.py

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
# -*- encoding=utf-8 -*-
 
# 导入包
import cv2
from functools import reduce
from PIL import Image
 
 
# 计算两个图片相似度函数ORB算法
def ORB_img_similarity(img1_path, img2_path):
    """
    :param img1_path: 图片1路径
    :param img2_path: 图片2路径
    :return: 图片相似度
    """
    try:
        # 读取图片
        img1 = cv2.imread(img1_path, cv2.IMREAD_GRAYSCALE)
        img2 = cv2.imread(img2_path, cv2.IMREAD_GRAYSCALE)
 
        # 初始化ORB检测器
        orb = cv2.ORB_create()
        kp1, des1 = orb.detectAndCompute(img1, None)
        kp2, des2 = orb.detectAndCompute(img2, None)
 
        # 提取并计算特征点
        bf = cv2.BFMatcher(cv2.NORM_HAMMING)
        # knn筛选结果
        matches = bf.knnMatch(des1, trainDescriptors=des2, k=2)
 
        # 查看最大匹配点数目
        good = [m for (m, n) in matches if m.distance < 0.75 * n.distance]
        similary = len(good) / len(matches)
        return similary
 
    except:
        return '0'
 
 
# 计算图片的局部哈希值--pHash
def phash(img):
    """
    :param img: 图片
    :return: 返回图片的局部hash值
    """
    img = img.resize((8, 8), Image.ANTIALIAS).convert('L')
    avg = reduce(lambda x, y: x + y, img.getdata()) / 64.
    hash_value = reduce(lambda x, y: x | (y[1] << y[0]), enumerate(map(lambda i: 0 if i < avg else 1, img.getdata())),
                        0)
    return hash_value
 
 
# 计算两个图片相似度函数局部敏感哈希算法
def phash_img_similarity(img1_path, img2_path):
    """
    :param img1_path: 图片1路径
    :param img2_path: 图片2路径
    :return: 图片相似度
    """
    # 读取图片
    img1 = Image.open(img1_path)
    img2 = Image.open(img2_path)
 
    # 计算汉明距离
    distance = bin(phash(img1) ^ phash(img2)).count('1')
    similary = 1 - distance / max(len(bin(phash(img1))), len(bin(phash(img1))))
    return similary
 
 
# 直方图计算图片相似度算法
def make_regalur_image(img, size=(256, 256)):
    """我们有必要把所有的图片都统一到特别的规格,在这里我选择是的256x256的分辨率。"""
    return img.resize(size).convert('RGB')
 
 
def hist_similar(lh, rh):
    assert len(lh) == len(rh)
    return sum(1 - (0 if l == r else float(abs(l - r)) / max(l, r)) for l, r in zip(lh, rh)) / len(lh)
 
 
def calc_similar(li, ri):
    return sum(hist_similar(l.histogram(), r.histogram()) for l, r in zip(split_image(li), split_image(ri))) / 16.0
 
 
def calc_similar_by_path(lf, rf):
    li, ri = make_regalur_image(Image.open(lf)), make_regalur_image(Image.open(rf))
    return calc_similar(li, ri)
 
 
def split_image(img, part_size=(64, 64)):
    w, h = img.size
    pw, ph = part_size
    assert w % pw == h % ph == 0
    return [img.crop((i, j, i + pw, j + ph)).copy() for i in range(0, w, pw) \
            for j in range(0, h, ph)]

五、结果

Python图片检索之以图搜图

到此这篇关于Python图片检索之以图搜图的文章就介绍到这了,更多相关Python以图搜图内容请搜索服务器之家以前的文章或继续浏览下面的相关文章希望大家以后多多支持服务器之家!

原文链接:https://blog.csdn.net/weixin_43723625/article/details/117298412

延伸 · 阅读

精彩推荐