[Python] opencvで画像の類似度を求める方法

import cv2
# 画像を読み込む
image = cv2.imread('path/to/image.jpg')
# 画像を表示する
cv2.imshow('Loaded Image', image)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
import numpy as np
import matplotlib.pyplot as plt
# 画像を読み込む
image = cv2.imread('path/to/image.jpg')
# ヒストグラムを計算する
hist = cv2.calcHist([image], [0], None, [256], [0, 256])
# ヒストグラムを表示する
plt.plot(hist)
plt.title('Histogram')
plt.xlabel('Pixel Value')
plt.ylabel('Frequency')
plt.show()

import cv2
# 2つの画像を読み込む
image1 = cv2.imread('path/to/image1.jpg')
image2 = cv2.imread('path/to/image2.jpg')
# ヒストグラムを計算する
hist1 = cv2.calcHist([image1], [0], None, [256], [0, 256])
hist2 = cv2.calcHist([image2], [0], None, [256], [0, 256])
# ヒストグラムを比較する
similarity = cv2.compareHist(hist1, hist2, cv2.HISTCMP_CORREL)
print(f'ヒストグラムの類似度: {similarity}')

ヒストグラムの類似度: 0.9967426773549225

import cv2
from skimage.metrics import structural_similarity as ssim
# 画像を読み込む
image1 = cv2.imread('path/to/image1.jpg', cv2.IMREAD_GRAYSCALE)
image2 = cv2.imread('path/to/image2.jpg', cv2.IMREAD_GRAYSCALE)
# SSIMを計算する
similarity_index, _ = ssim(image1, image2, full=True)
print(f'SSIMの類似度: {similarity_index}')

import cv2
from skimage.metrics import structural_similarity as ssim
# 画像を読み込む
image1 = cv2.imread('path/to/image1.jpg', cv2.IMREAD_GRAYSCALE)
image2 = cv2.imread('path/to/image2.jpg', cv2.IMREAD_GRAYSCALE)
# SSIMを計算する
similarity_index, _ = ssim(image1, image2, full=True)
# 結果を表示する
print(f'SSIMの類似度: {similarity_index}')

import cv2
# 画像を読み込む
image = cv2.imread('path/to/image.jpg')
# ORBオブジェクトを作成
orb = cv2.ORB_create()
# 特徴点を検出
keypoints, descriptors = orb.detectAndCompute(image, None)
# 特徴点を描画
image_with_keypoints = cv2.drawKeypoints(image, keypoints, None, color=(0, 255, 0))
# 結果を表示
cv2.imshow('Keypoints', image_with_keypoints)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
# 2つの画像を読み込む
image1 = cv2.imread('path/to/image1.jpg')
image2 = cv2.imread('path/to/image2.jpg')
# ORBオブジェクトを作成
orb = cv2.ORB_create()
# 特徴点を検出
keypoints1, descriptors1 = orb.detectAndCompute(image1, None)
keypoints2, descriptors2 = orb.detectAndCompute(image2, None)
# BFMatcherを作成
bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
# 特徴量をマッチング
matches = bf.match(descriptors1, descriptors2)
# マッチング結果を描画
matches = sorted(matches, key=lambda x: x.distance)
matched_image = cv2.drawMatches(image1, keypoints1, image2, keypoints2, matches[:10], None, flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)
# 結果を表示
cv2.imshow('Matches', matched_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
# 2つの画像を読み込む
image1 = cv2.imread('path/to/image1.jpg')
image2 = cv2.imread('path/to/image2.jpg')
# ORBオブジェクトを作成
orb = cv2.ORB_create()
# 特徴点を検出
keypoints1, descriptors1 = orb.detectAndCompute(image1, None)
keypoints2, descriptors2 = orb.detectAndCompute(image2, None)
# BFMatcherを作成
bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
# 特徴量をマッチング
matches = bf.match(descriptors1, descriptors2)
# マッチング結果を描画
matches = sorted(matches, key=lambda x: x.distance)
matched_image = cv2.drawMatches(image1, keypoints1, image2, keypoints2, matches[:10], None, flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)
# 結果を表示
cv2.imshow('Matches', matched_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
import numpy as np
# 対象画像とテンプレート画像を読み込む
image = cv2.imread('path/to/image.jpg')
template = cv2.imread('path/to/template.jpg')
# テンプレートマッチングを実行
result = cv2.matchTemplate(image, template, cv2.TM_CCOEFF_NORMED)
# 類似度が最も高い位置を特定
min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(result)
# テンプレートのサイズを取得
h, w = template.shape[:2]
# マッチング結果を描画
top_left = max_loc
bottom_right = (top_left[0] + w, top_left[1] + h)
cv2.rectangle(image, top_left, bottom_right, (0, 255, 0), 2)
# 結果を表示
cv2.imshow('Matched Result', image)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
import numpy as np
# 対象画像とテンプレート画像を読み込む
image = cv2.imread('path/to/image.jpg')
template = cv2.imread('path/to/template.jpg')
# テンプレートマッチングを実行
result = cv2.matchTemplate(image, template, cv2.TM_CCOEFF_NORMED)
# 類似度が最も高い位置を特定
min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(result)
# テンプレートのサイズを取得
h, w = template.shape[:2]
# マッチング結果を描画
top_left = max_loc
bottom_right = (top_left[0] + w, top_left[1] + h)
cv2.rectangle(image, top_left, bottom_right, (0, 255, 0), 2)
# 結果を表示
cv2.imshow('Matched Result', image)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
import numpy as np
import os
def calculate_histogram(image_path):
    image = cv2.imread(image_path)
    hist = cv2.calcHist([image], [0], None, [256], [0, 256])
    return hist
def search_similar_images(query_image_path, database_folder):
    query_hist = calculate_histogram(query_image_path)
    similarities = {}
    for filename in os.listdir(database_folder):
        if filename.endswith('.jpg') or filename.endswith('.png'):
            db_image_path = os.path.join(database_folder, filename)
            db_hist = calculate_histogram(db_image_path)
            similarity = cv2.compareHist(query_hist, db_hist, cv2.HISTCMP_CORREL)
            similarities[filename] = similarity
    # 類似度が高い順にソート
    sorted_similarities = sorted(similarities.items(), key=lambda x: x[1], reverse=True)
    return sorted_similarities
# 使用例
query_image = 'path/to/query_image.jpg'
database_folder = 'path/to/image_database/'
results = search_similar_images(query_image, database_folder)
for filename, similarity in results:
    print(f'画像: {filename}, 類似度: {similarity}')

import cv2
from skimage.metrics import structural_similarity as ssim
import os
def calculate_ssim(image_path1, image_path2):
    image1 = cv2.imread(image_path1, cv2.IMREAD_GRAYSCALE)
    image2 = cv2.imread(image_path2, cv2.IMREAD_GRAYSCALE)
    similarity_index, _ = ssim(image1, image2, full=True)
    return similarity_index
def search_similar_images_ssim(query_image_path, database_folder):
    similarities = {}
    for filename in os.listdir(database_folder):
        if filename.endswith('.jpg') or filename.endswith('.png'):
            db_image_path = os.path.join(database_folder, filename)
            similarity = calculate_ssim(query_image_path, db_image_path)
            similarities[filename] = similarity
    # 類似度が高い順にソート
    sorted_similarities = sorted(similarities.items(), key=lambda x: x[1], reverse=True)
    return sorted_similarities
# 使用例
query_image = 'path/to/query_image.jpg'
database_folder = 'path/to/image_database/'
results = search_similar_images_ssim(query_image, database_folder)
for filename, similarity in results:
    print(f'画像: {filename}, SSIM類似度: {similarity}')

import cv2
import os
def calculate_keypoints_and_descriptors(image_path):
    image = cv2.imread(image_path)
    orb = cv2.ORB_create()
    keypoints, descriptors = orb.detectAndCompute(image, None)
    return keypoints, descriptors
def search_similar_images_feature_matching(query_image_path, database_folder):
    query_keypoints, query_descriptors = calculate_keypoints_and_descriptors(query_image_path)
    bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
    similarities = {}
    for filename in os.listdir(database_folder):
        if filename.endswith('.jpg') or filename.endswith('.png'):
            db_image_path = os.path.join(database_folder, filename)
            db_keypoints, db_descriptors = calculate_keypoints_and_descriptors(db_image_path)
            matches = bf.match(query_descriptors, db_descriptors)
            similarity = len(matches)  # マッチした特徴点の数を類似度とする
            similarities[filename] = similarity
    # 類似度が高い順にソート
    sorted_similarities = sorted(similarities.items(), key=lambda x: x[1], reverse=True)
    return sorted_similarities
# 使用例
query_image = 'path/to/query_image.jpg'
database_folder = 'path/to/image_database/'
results = search_similar_images_feature_matching(query_image, database_folder)
for filename, similarity in results:
    print(f'画像: {filename}, マッチした特徴点数: {similarity}')

import cv2
import numpy as np
import os
def calculate_histogram(image_path):
    image = cv2.imread(image_path)
    hist = cv2.calcHist([image], [0], None, [256], [0, 256])
    return hist
def find_duplicate_images(database_folder):
    histograms = {}
    duplicates = []
    for filename in os.listdir(database_folder):
        if filename.endswith('.jpg') or filename.endswith('.png'):
            image_path = os.path.join(database_folder, filename)
            hist = calculate_histogram(image_path)
            histograms[filename] = hist
    # ヒストグラムを比較して重複を検出
    for file1, hist1 in histograms.items():
        for file2, hist2 in histograms.items():
            if file1 != file2:
                similarity = cv2.compareHist(hist1, hist2, cv2.HISTCMP_CORREL)
                if similarity > 0.95:  # 類似度が高い場合
                    duplicates.append((file1, file2))
    return duplicates
# 使用例
database_folder = 'path/to/image_database/'
duplicates = find_duplicate_images(database_folder)
for file1, file2 in duplicates:
    print(f'重複画像: {file1} と {file2}')

import cv2
from skimage.metrics import structural_similarity as ssim
import os
def calculate_ssim(image_path1, image_path2):
    image1 = cv2.imread(image_path1, cv2.IMREAD_GRAYSCALE)
    image2 = cv2.imread(image_path2, cv2.IMREAD_GRAYSCALE)
    similarity_index, _ = ssim(image1, image2, full=True)
    return similarity_index
def find_duplicate_images_ssim(database_folder):
    duplicates = []
    images = [f for f in os.listdir(database_folder) if f.endswith('.jpg') or f.endswith('.png')]
    for i in range(len(images)):
        for j in range(i + 1, len(images)):
            image_path1 = os.path.join(database_folder, images[i])
            image_path2 = os.path.join(database_folder, images[j])
            similarity = calculate_ssim(image_path1, image_path2)
            if similarity > 0.95:  # 類似度が高い場合
                duplicates.append((images[i], images[j]))
    return duplicates
# 使用例
database_folder = 'path/to/image_database/'
duplicates = find_duplicate_images_ssim(database_folder)
for file1, file2 in duplicates:
    print(f'重複画像: {file1} と {file2}')

import cv2
import os
def calculate_keypoints_and_descriptors(image_path):
    image = cv2.imread(image_path)
    orb = cv2.ORB_create()
    keypoints, descriptors = orb.detectAndCompute(image, None)
    return keypoints, descriptors
def find_duplicate_images_feature_matching(database_folder):
    duplicates = []
    images = [f for f in os.listdir(database_folder) if f.endswith('.jpg') or f.endswith('.png')]
    for i in range(len(images)):
        keypoints1, descriptors1 = calculate_keypoints_and_descriptors(os.path.join(database_folder, images[i]))
        for j in range(i + 1, len(images)):
            keypoints2, descriptors2 = calculate_keypoints_and_descriptors(os.path.join(database_folder, images[j]))
            bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
            matches = bf.match(descriptors1, descriptors2)
            if len(matches) > 10:  # マッチした特徴点が一定数以上の場合
                duplicates.append((images[i], images[j]))
    return duplicates
# 使用例
database_folder = 'path/to/image_database/'
duplicates = find_duplicate_images_feature_matching(database_folder)
for file1, file2 in duplicates:
    print(f'重複画像: {file1} と {file2}')

import cv2
import numpy as np
def calculate_histogram(frame):
    hist = cv2.calcHist([frame], [0], None, [256], [0, 256])
    return hist
def compare_frames_histogram(video_path):
    cap = cv2.VideoCapture(video_path)
    ret, prev_frame = cap.read()
    prev_hist = calculate_histogram(prev_frame)
    frame_index = 1
    similarities = []
    while True:
        ret, current_frame = cap.read()
        if not ret:
            break
        current_hist = calculate_histogram(current_frame)
        similarity = cv2.compareHist(prev_hist, current_hist, cv2.HISTCMP_CORREL)
        similarities.append((frame_index, similarity))
        prev_hist = current_hist
        frame_index += 1
    cap.release()
    return similarities
# 使用例
video_path = 'path/to/video.mp4'
similarities = compare_frames_histogram(video_path)
for index, similarity in similarities:
    print(f'フレーム {index}: 類似度 {similarity}')

import cv2
from skimage.metrics import structural_similarity as ssim
def compare_frames_ssim(video_path):
    cap = cv2.VideoCapture(video_path)
    ret, prev_frame = cap.read()
    prev_frame_gray = cv2.cvtColor(prev_frame, cv2.COLOR_BGR2GRAY)
    frame_index = 1
    similarities = []
    while True:
        ret, current_frame = cap.read()
        if not ret:
            break
        current_frame_gray = cv2.cvtColor(current_frame, cv2.COLOR_BGR2GRAY)
        similarity_index, _ = ssim(prev_frame_gray, current_frame_gray, full=True)
        similarities.append((frame_index, similarity_index))
        prev_frame_gray = current_frame_gray
        frame_index += 1
    cap.release()
    return similarities
# 使用例
video_path = 'path/to/video.mp4'
similarities = compare_frames_ssim(video_path)
for index, similarity in similarities:
    print(f'フレーム {index}: SSIM類似度 {similarity}')

import cv2
def compare_frames_feature_matching(video_path):
    cap = cv2.VideoCapture(video_path)
    orb = cv2.ORB_create()
    ret, prev_frame = cap.read()
    prev_keypoints, prev_descriptors = orb.detectAndCompute(prev_frame, None)
    frame_index = 1
    similarities = []
    while True:
        ret, current_frame = cap.read()
        if not ret:
            break
        current_keypoints, current_descriptors = orb.detectAndCompute(current_frame, None)
        bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
        matches = bf.match(prev_descriptors, current_descriptors)
        similarity = len(matches)  # マッチした特徴点の数を類似度とする
        similarities.append((frame_index, similarity))
        prev_keypoints, prev_descriptors = current_keypoints, current_descriptors
        frame_index += 1
    cap.release()
    return similarities
# 使用例
video_path = 'path/to/video.mp4'
similarities = compare_frames_feature_matching(video_path)
for index, similarity in similarities:
    print(f'フレーム {index}: マッチした特徴点数 {similarity}')