感知哈希算法计算图片相似度

from PIL import Image
import numpy as np
import os
import pandas as pd


def image_to_grayscale(image_path):
    """将图像转换为灰度图"""
    img = Image.open(image_path).convert('L')
    return img


def reduce_image_size(img, size=8):
    """将图像缩小到指定大小"""
    img = img.resize((size, size), Image.ANTIALIAS)
    return img


def compute_average_hash(img):
    """计算图像的平均哈希值"""
    pixels = np.array(img, dtype=np.float32)
    avg = pixels.mean()
    diff = pixels > avg
    hash_ = ''.join(map(str, diff.flatten().astype(int)))
    return hash_


def phash(image_path, hash_size=8):
    """计算图像的pHash值"""
    img = image_to_grayscale(image_path)
    img = reduce_image_size(img, hash_size)
    return compute_average_hash(img)


def calculate_image_similarities(folder_path):
    """计算文件夹中所有图片两两之间的相似度"""
    images = [os.path.join(folder_path, f) for f in os.listdir(folder_path) if
              f.lower().endswith(('.png', '.jpg', '.jpeg'))]
    n_images = len(images)
    similarity_matrix = np.zeros((n_images, n_images), dtype=int)

    for i, img_path1 in enumerate(images):
        hash1 = phash(img_path1)
        for j, img_path2 in enumerate(images[i + 1:]):  # 避免重复计算并节省时间
            hash2 = phash(img_path2)
            distance = sum(c1 != c2 for c1, c2 in zip(hash1, hash2))
            similarity_matrix[i, j + i + 1] = distance
            similarity_matrix[j + i + 1, i] = distance  # 对称矩阵

    return similarity_matrix, images


def generate_similarity_table(similarity_matrix, images):
    """生成相似度表格"""
    df = pd.DataFrame(similarity_matrix, index=images, columns=images)
    return df


# 示例使用
folder_path = 'C:\\Users\\Desktop\\全部图片数据'
similarity_matrix, images = calculate_image_similarities(folder_path)
similarity_table = generate_similarity_table(similarity_matrix, images)

# 显示表格（可能需要调整显示设置以适合大量数据）
print(similarity_table.head())  # 仅显示前几行，因为整个表格可能非常大

# 保存表格到CSV文件
similarity_table.to_csv('image_similarities.csv')