import os
import random
import cv2
import numpy as np
from PIL import Image, ImageEnhance, ImageOps

# 定義各種數(shù)據(jù)增強(qiáng)方法
def random_rotate(image, angle_range=(-30, 30)):
    angle = random.uniform(angle_range[0], angle_range[1])
    (h, w) = image.shape[:2]
    center = (w // 2, h // 2)
    M = cv2.getRotationMatrix2D(center, angle, 1.0)
    rotated = cv2.warpAffine(image, M, (w, h), borderMode=cv2.BORDER_REFLECT)
    return rotated

def random_translate(image, translate_range=(-50, 50)):
    tx = random.randint(translate_range[0], translate_range[1])
    ty = random.randint(translate_range[0], translate_range[1])
    (h, w) = image.shape[:2]
    M = np.float32([[1, 0, tx], [0, 1, ty]])
    translated = cv2.warpAffine(image, M, (w, h), borderMode=cv2.BORDER_REFLECT)
    return translated

def random_flip(image):
    flip_code = random.choice([-1, 0, 1])
    flipped = cv2.flip(image, flip_code)
    return flipped

def random_scale(image, scale_range=(0.8, 1.2)):
    scale = random.uniform(scale_range[0], scale_range[1])
    (h, w) = image.shape[:2]
    new_dim = (int(w * scale), int(h * scale))
    scaled = cv2.resize(image, new_dim, interpolation=cv2.INTER_LINEAR)
    return scaled

def random_crop(image, crop_size=(224, 224)):
    (h, w) = image.shape[:2]
    if crop_size[0] > h or crop_size[1] > w:
        # 當(dāng)裁剪尺寸大于圖像尺寸時，拋出異?；蛘{(diào)整裁剪尺寸
        raise ValueError("Crop size is larger than image size.")
    top = random.randint(0, h - crop_size[0])
    left = random.randint(0, w - crop_size[1])
    cropped = image[top:top+crop_size[0], left:left+crop_size[1]]
    return cropped

def random_color_jitter(image):
    pil_image = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    color_jitter = ImageEnhance.Color(pil_image).enhance(random.uniform(0.6, 1.4))
    contrast_jitter = ImageEnhance.Contrast(color_jitter).enhance(random.uniform(0.5, 1.5))
    brightness_jitter = ImageEnhance.Brightness(contrast_jitter).enhance(random.uniform(0.6, 1.4))
    sharpness_jitter = ImageEnhance.Sharpness(brightness_jitter).enhance(random.uniform(0.6, 1.4))
    jittered = cv2.cvtColor(np.array(sharpness_jitter), cv2.COLOR_RGB2BGR)
    return jittered

def random_add_noise(image):
    row, col, ch = image.shape
    mean = 0
    var = 0.1
    sigma = var ** 0.5
    gauss = np.random.normal(mean, sigma, (row, col, ch))
    gauss = gauss.reshape(row, col, ch)
    noisy = image + gauss
    return np.clip(noisy, 0, 255).astype(np.uint8)

# 數(shù)據(jù)增強(qiáng)主函數(shù)
def augment_random_images(src_folder, dst_folder, num_images_to_select, num_augmentations_per_image):
    if not os.path.exists(dst_folder):
        os.makedirs(dst_folder)

    # 獲取所有圖像文件名
    all_filenames = [f for f in os.listdir(src_folder) if f.lower().endswith(('.png', '.jpg', '.jpeg'))]

    # 如果選擇的圖像數(shù)量大于總圖像數(shù)量，則只處理全部圖像
    num_images_to_process = min(num_images_to_select, len(all_filenames))

    # 隨機(jī)選擇圖像
    selected_filenames = random.sample(all_filenames, num_images_to_process)

    # 創(chuàng)建一個增強(qiáng)方法列表
    augmentation_methods = [
        random_rotate,
        #random_translate,
        random_flip,
        random_scale,
        #random_crop,
        random_color_jitter,
        random_add_noise
    ]

    for filename in selected_filenames:
        img_path = os.path.join(src_folder, filename)
        image = cv2.imread(img_path)

        for i in range(num_augmentations_per_image):
            # 隨機(jī)選擇一種增強(qiáng)方法
            augmentation_method = random.choice(augmentation_methods)
            
            # 應(yīng)用選中的增強(qiáng)方法
            augmented_img = augmentation_method(image)

            # 保存增強(qiáng)后的圖像
            base_name, ext = os.path.splitext(filename)
            save_path = os.path.join(dst_folder, f"{base_name}_aug_{i}{ext}")
            cv2.imwrite(save_path, augmented_img)

if __name__ == "__main__":
    src_folder = 'path/to/source/folder'  # 替換為你的源文件夾路徑
    dst_folder = 'path/to/destination/folder'  # 替換為你要保存增強(qiáng)圖像的文件夾路徑
    num_images_to_select = 10  # 從源文件夾中隨機(jī)選擇的圖像數(shù)量
    num_augmentations_per_image = 5  # 每張圖像生成的增強(qiáng)圖像數(shù)量

    augment_random_images(src_folder, dst_folder, num_images_to_select, num_augmentations_per_image)
    print(f"圖像增強(qiáng)完成，增強(qiáng)后的圖像已保存到 {dst_folder}")