要避免从PDF中提取重复的图片，可以使用以下解决方法：

1. 使用图像哈希算法：使用图像哈希算法（如pHash或dHash）对提取的图片进行哈希处理。然后，比较图像哈希值，如果两个图像的哈希值相似，则可以认为它们是重复的图片。

import cv2
import imagehash

def extract_image_hashes(pdf_file):
    image_hashes = []
    with pdfplumber.open(pdf_file) as pdf:
        for page in pdf.pages:
            for image in page.images:
                img = image.to_image()
                img_cv2 = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
                img_hash = imagehash.phash(Image.fromarray(img_cv2))
                image_hashes.append(img_hash)
    return image_hashes

def remove_duplicate_images(pdf_file):
    image_hashes = extract_image_hashes(pdf_file)
    unique_image_hashes = list(set(image_hashes))
    
    with pdfplumber.open(pdf_file) as pdf:
        for page in pdf.pages:
            for image in page.images:
                img = image.to_image()
                img_cv2 = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
                img_hash = imagehash.phash(Image.fromarray(img_cv2))
                if img_hash in unique_image_hashes:
                    unique_image_hashes.remove(img_hash)
                else:
                    # save the unique image or perform some other action

2. 使用图像特征向量：使用深度学习模型（如VGG16、ResNet等）提取图像的特征向量。然后，通过比较特征向量的相似度来判断图像是否重复。

import cv2
import numpy as np
from keras.applications.vgg16 import VGG16
from keras.applications.vgg16 import preprocess_input
from keras.models import Model

def extract_image_features(pdf_file):
    image_features = []
    base_model = VGG16(weights='imagenet')
    model = Model(inputs=base_model.input, outputs=base_model.get_layer('fc1').output)
    
    with pdfplumber.open(pdf_file) as pdf:
        for page in pdf.pages:
            for image in page.images:
                img = image.to_image()
                img_cv2 = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
                img_resized = cv2.resize(img_cv2, (224, 224))
                img_preprocessed = preprocess_input(np.expand_dims(img_resized, axis=0))
                img_feature = model.predict(img_preprocessed)[0]
                image_features.append(img_feature)
    return image_features

def remove_duplicate_images(pdf_file):
    image_features = extract_image_features(pdf_file)
    unique_image_features = []
    
    with pdfplumber.open(pdf_file) as pdf:
        for page in pdf.pages:
            for image in page.images:
                img = image.to_image()
                img_cv2 = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
                img_resized = cv2.resize(img_cv2, (224, 224))
                img_preprocessed = preprocess_input(np.expand_dims(img_resized, axis=0))
                img_feature = model.predict(img_preprocessed)[0]
                
                if not any(np.allclose(img_feature, feat) for feat in unique_image_features):
                    unique_image_features.append(img_feature)
                else:
                    # save the unique image or perform some other action

这些示例代码使用了Python的pdfplumber库来从PDF中提取图片，并使用了OpenCV和PIL库来处理图像。第一个示例使用了图像哈希算法来进行判断，而第二个示例使用了深度学习模型来提取图像的特征向量。根据具体情况选择适合的方法。