|
|
import cv2
|
|
|
import numpy as np
|
|
|
import PIL.Image as Image
|
|
|
import torchvision.transforms as transforms
|
|
|
|
|
|
class CLAHE:
|
|
|
def __init__(self, clipLimit=2.0, tileGridSize=(8, 8)):
|
|
|
# 创建 CLAHE 对象
|
|
|
self.clipLimit = clipLimit
|
|
|
self.tileGridSize = tileGridSize
|
|
|
|
|
|
def __call__(self, img):
|
|
|
# 将 PIL 图像转换为 NumPy 数组
|
|
|
img_np = np.array(img)
|
|
|
|
|
|
# 创建 CLAHE 对象（推迟到 __call__ 方法中创建）
|
|
|
clahe = cv2.createCLAHE(clipLimit=self.clipLimit, tileGridSize=self.tileGridSize)
|
|
|
|
|
|
# 应用 CLAHE
|
|
|
img_clahe = clahe.apply(img_np)
|
|
|
|
|
|
# 将 NumPy 数组转换回 PIL 图像
|
|
|
img_clahe_pil = Image.fromarray(img_clahe)
|
|
|
|
|
|
return img_clahe_pil
|
|
|
|
|
|
def TrainCompose() -> transforms.Compose:
|
|
|
return transforms.Compose(
|
|
|
[
|
|
|
transforms.Grayscale(), # 将图片转换为单通道灰度图
|
|
|
transforms.RandomVerticalFlip(p=0.5), # 随机垂直翻转图片
|
|
|
transforms.RandomHorizontalFlip(p=0.5), # 随机水平翻转图片
|
|
|
transforms.RandomRotation(degrees=(-20, 20)), # 随机对图片进行旋转
|
|
|
CLAHE(), # 对图片进行clahe处理
|
|
|
transforms.ToTensor(), # 将图片转换为Tensor，同时把图片的数据从[0, 255]转换到[0, 1]
|
|
|
transforms.Normalize(mean=(0.5,), std=(0.5,)), # 归一化到[-1,1]，均值和标准差都是0.5(可调)
|
|
|
transforms.Resize(size=(956, 956), antialias=True), # 将图片缩放
|
|
|
]
|
|
|
)
|
|
|
|
|
|
def TestCompose() -> transforms.Compose:
|
|
|
return transforms.Compose(
|
|
|
[
|
|
|
transforms.Grayscale(), # 将图片转换为单通道灰度图
|
|
|
CLAHE(), # 对图片进行clahe处理
|
|
|
transforms.ToTensor(), # 将图片转换为Tensor，同时把图片的数据从[0, 255]转换到[0, 1]
|
|
|
transforms.Normalize(mean=(0.5,), std=(0.5,)), # 归一化到[-1,1]，均值和标准差都是0.5(可调)
|
|
|
transforms.Resize(size=(956, 956), antialias=True), # 将图片缩放
|
|
|
]
|
|
|
) |
