有些数据集给的掩码是全黑图片，需要将全黑的掩码剔除，保留有标签的掩码。

DDR-dataset 眼底图像处理

 

from PIL import Image
import cv2
import os

def extract_mask_and_original(mask_path, original_path, output_folder):
    # 读取黑白掩码图片和同名原图图片
    mask_image = Image.open(mask_path).convert("L")
    original_image = Image.open(original_path)

    has_white_pixel = any(pixel > 0 for pixel in mask_image.getdata())

    if has_white_pixel:
        # 提取同名的原图图片
        original_filename = os.path.basename(mask_path).replace(".tif", ".jpg")
        original_path = os.path.join(input_image_folder, original_filename)

        # 保存掩码图片和原图图片
        output_mask_path = os.path.join(output_folder_mask, os.path.basename(mask_path))
        output_original_path = os.path.join(output_folder_image, original_filename)

        mask_image.save(output_mask_path)
        original_image.save(output_original_path)

        print(f"已提取掩码和原图：{output_mask_path}，{output_original_path}")
    else:
        print(f"掩码中不存在白色像素：{mask_path}")

    print("提取完成")


# 设置输入文件夹和输出文件夹
input_mask_folder = r"E:\yanjiu\datasets\DDR_segmentation\valid\mask"
input_image_folder = r"E:\yanjiu\datasets\DDR_segmentation\valid\image"
output_folder_mask = r"E:\yanjiu\datasets\DDR_segmentation\valid\mask_new"
output_folder_image =r"E:\yanjiu\datasets\DDR_segmentation\valid\image_new"

# 遍历输入文件夹中的掩码图片
for mask_filename in os.listdir(input_mask_folder):
    if mask_filename.endswith(".tif"):
        mask_path = os.path.join(input_mask_folder, mask_filename)
        original_filename = mask_filename.replace(".tif", ".jpg")
        original_path = os.path.join(input_image_folder, original_filename)

        extract_mask_and_original(mask_path, original_path, output_folder_mask)