| 1 | import os |
| 2 | import requests |
| 3 | import numpy as np |
| 4 | from pathlib import Path |
| 5 | from datetime import datetime, timedelta |
| 6 | from PIL import Image |
| 7 | import logging |
| 8 | |
| 9 | |
| 10 | class ImageProcessor: |
| 11 | """ |
| 12 | 图片处理和上传集合类 |
| 13 | 支持：图片处理、分割、拼接，以及上传到阿里云OSS |
| 14 | """ |
| 15 | |
| 16 | # 阿里云DashScope上传配置 |
| 17 | UPLOAD_API_URL = "https://dashscope.aliyuncs.com/api/v1/uploads" |
| 18 | |
| 19 | def __init__(self, |
| 20 | image_path='', |
| 21 | api_key: str = "sk-bcab316d69a7414faa9dc29737019333", |
| 22 | model_name: str = "wan2.6-i2v-flash", |
| 23 | local_proxy: str | None = None): |
| 24 | """ |
| 25 | 初始化图片处理器 |
| 26 | |
| 27 | Args: |
| 28 | image_path: 图片文件路径（可选，用于处理已有图片） |
| 29 | api_key: DashScope API Key（用于上传，可从环境变量 DASHSCOPE_API_KEY 读取） |
| 30 | model_name: 模型名称，默认使用 wan2.6-i2v-flash |
| 31 | """ |
| 32 | # 图片处理部分 |
| 33 | if image_path != '': |
| 34 | self.image_path = image_path |
| 35 | self.image = Image.open(image_path) |
| 36 | self.image_np = np.array(self.image) |
| 37 | self.width, self.height = self.image_np.shape[1], self.image_np.shape[0] |
| 38 | else: |
| 39 | self.image_path = None |
| 40 | self.image = None |
| 41 | self.image_np = None |
| 42 | self.width = None |
| 43 | self.height = None |
| 44 | |
| 45 | # 上传功能部分 |
| 46 | self.api_key = api_key or os.getenv("DASHSCOPE_API_KEY") |
| 47 | self.model_name = model_name |
| 48 | self.local_proxy = local_proxy |
| 49 | |
| 50 | def _proxies(self): |
| 51 | if not self.local_proxy: |
| 52 | return None |
| 53 | return {"http": self.local_proxy, "https": self.local_proxy} |
| 54 | |
| 55 | @staticmethod |
| 56 | def check_column_white(column_pixels): |
| 57 | """检查列是否几乎全白""" |
| 58 | is_almost_white = np.logical_or(column_pixels == 254, column_pixels == 255) |
| 59 | white_pixels_ratio = np.mean(np.all(is_almost_white, axis=-1)) |
| 60 | return white_pixels_ratio >= 0.98 # 至少98%的像素为白色 |
| 61 | |
| 62 | def find_white_section(self, start, end): |
| 63 | """查找指定范围内的白色区间""" |
| 64 | white_sections = [] |
| 65 | in_white_section = False |
| 66 | start_index = 0 |
| 67 | |
| 68 | for col in range(start, end): |
| 69 | column_pixels = self.image_np[:, col, :] |
| 70 | if self.check_column_white(column_pixels): |
| 71 | if not in_white_section: |
| 72 | start_index = col |
| 73 | in_white_section = True |
| 74 | else: |
| 75 | if in_white_section: |
| 76 | white_sections.append((start_index, col)) |
| 77 | in_white_section = False |
| 78 | |
| 79 | if in_white_section: |
| 80 | white_sections.append((start_index, end)) |
| 81 | |
| 82 | return white_sections |
| 83 | |
| 84 | def split_image(self): |
| 85 | """将图片从中间分割为左右两部分""" |
| 86 | start_col = self.width * 2 // 5 |
| 87 | end_col = self.width * 3 // 5 |
| 88 | white_sections = self.find_white_section(start_col, end_col) |
| 89 | |
| 90 | if white_sections: |
| 91 | middle_section = white_sections[len(white_sections) // 2] |
| 92 | mid_col = (middle_section[0] + middle_section[1]) // 2 |
| 93 | else: |
| 94 | raise ValueError("No suitable white column found within the specified range") |
| 95 | |
| 96 | left_box = (0, 0, mid_col, self.height) |
| 97 | right_box = (mid_col, 0, self.width, self.height) |
| 98 | left_image = self.image.crop(left_box) |
| 99 | right_image = self.image.crop(right_box) |
| 100 | |
| 101 | save_dir, filename = os.path.split(self.image_path) |
| 102 | base, extension = os.path.splitext(filename) |
| 103 | |
| 104 | left_image_path = os.path.join(save_dir, base + '_front' + extension) |
| 105 | right_image_path = os.path.join(save_dir, base + '_back' + extension) |
| 106 | left_image.save(left_image_path) |
| 107 | right_image.save(right_image_path) |
| 108 | |
| 109 | return left_image_path, right_image_path |
| 110 | |
| 111 | def stitch_images(self, image_paths, output_path): |
| 112 | """拼接多张图片""" |
| 113 | if not image_paths: |
| 114 | raise ValueError("No image paths provided") |
| 115 | sample_image = Image.open(image_paths[0]) |
| 116 | single_width, single_height = sample_image.size |
| 117 | num_images = len(image_paths) |
| 118 | total_desired_width = single_width |
| 119 | total_current_width = single_width * num_images |
| 120 | total_width_to_cut = max(0, total_current_width - total_desired_width) |
| 121 | width_to_cut_per_image = total_width_to_cut // num_images |
| 122 | stitched_image = Image.new('RGB', (total_desired_width, single_height), "white") |
| 123 | current_x = 0 |
| 124 | |
| 125 | for path in image_paths: |
| 126 | image = Image.open(path) |
| 127 | if width_to_cut_per_image > 0: |
| 128 | left_margin = width_to_cut_per_image // 2 |
| 129 | right_margin = image.width - width_to_cut_per_image + left_margin |
| 130 | image = image.crop((left_margin, 0, right_margin, image.height)) |
| 131 | stitched_image.paste(image, (current_x, 0)) |
| 132 | current_x += image.width |
| 133 | |
| 134 | output_dir = os.path.dirname(output_path) |
| 135 | if not os.path.exists(output_dir): |
| 136 | os.makedirs(output_dir) |
| 137 | stitched_image.save(output_path) |
| 138 | return output_path |
| 139 | |
| 140 | def download_image(self, image_url, save_path, max_retries=3): |
| 141 | """ |
| 142 | 下载图片，带有重试机制和SSL错误处理 |
| 143 | |
| 144 | Args: |
| 145 | image_url: 图片URL |
| 146 | save_path: 本地保存路径 |
| 147 | max_retries: 最大重试次数 |
| 148 | """ |
| 149 | import time |
| 150 | import urllib3 |
| 151 | |
| 152 | urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning) |
| 153 | |
| 154 | for attempt in range(max_retries): |
| 155 | try: |
| 156 | response = requests.get( |
| 157 | image_url, |
| 158 | timeout=(10, 30), |
| 159 | stream=True, |
| 160 | verify=True, |
| 161 | proxies=self._proxies(), |
| 162 | headers={ |
| 163 | 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36' |
| 164 | } |
| 165 | ) |
| 166 | |
| 167 | if response.status_code == 200: |
| 168 | with open(save_path, 'wb') as file: |
| 169 | for chunk in response.iter_content(chunk_size=8192): |
| 170 | if chunk: |
| 171 | file.write(chunk) |
| 172 | print(f"✓ 图片下载成功: {save_path}") |
| 173 | return True |
| 174 | else: |
| 175 | print(f"下载失败，状态码: {response.status_code}") |
| 176 | |
| 177 | except requests.exceptions.SSLError as e: |
| 178 | print(f"SSL错误 (尝试 {attempt + 1}/{max_retries}): {str(e)[:100]}") |
| 179 | if attempt < max_retries - 1: |
| 180 | wait_time = (attempt + 1) * 2 |
| 181 | print(f"等待 {wait_time} 秒后重试...") |
| 182 | time.sleep(wait_time) |
| 183 | else: |
| 184 | print("尝试禁用SSL验证重新下载...") |
| 185 | try: |
| 186 | response = requests.get( |
| 187 | image_url, |
| 188 | timeout=(10, 30), |
| 189 | stream=True, |
| 190 | verify=False, |
| 191 | proxies=self._proxies(), |
| 192 | headers={ |
| 193 | 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36' |
| 194 | } |
| 195 | ) |
| 196 | if response.status_code == 200: |
| 197 | with open(save_path, 'wb') as file: |
| 198 | for chunk in response.iter_content(chunk_size=8192): |
| 199 | if chunk: |
| 200 | file.write(chunk) |
| 201 | print(f"✓ 图片下载成功(已禁用SSL验证): {save_path}") |
| 202 | return True |
| 203 | except Exception as fallback_error: |
| 204 | print(f"禁用SSL验证后仍然失败: {fallback_error}") |
| 205 | raise |
| 206 | |
| 207 | except requests.exceptions.Timeout as e: |
| 208 | print(f"超时错误 (尝试 {attempt + 1}/{max_retries}): {e}") |
| 209 | if attempt < max_retries - 1: |
| 210 | time.sleep((attempt + 1) * 2) |
| 211 | else: |
| 212 | raise |
| 213 | |
| 214 | except Exception as e: |
| 215 | print(f"下载错误 (尝试 {attempt + 1}/{max_retries}): {e}") |
| 216 | if attempt < max_retries - 1: |
| 217 | time.sleep((attempt + 1) * 2) |
| 218 | else: |
| 219 | raise |
| 220 | |
| 221 | return False |
| 222 | |
| 223 | def resize_image(self, image_path): |
| 224 | """调整图片大小（添加顶部空白）""" |
| 225 | original_image = Image.open(image_path) |
| 226 | width, height = original_image.size |
| 227 | top_blank_height = height // 2 |
| 228 | final_height = height + top_blank_height |
| 229 | final_width = int(final_height * 5 / 3) |
| 230 | new_image = Image.new("RGB", (final_width, final_height), color="white") |
| 231 | left = (final_width - width) // 2 |
| 232 | top = top_blank_height |
| 233 | new_image.paste(original_image, (left, top)) |
| 234 | new_image.save(image_path) |
| 235 | return image_path |
| 236 | |
| 237 | def has_black_borders(self, image_path, threshold=10, black_limit=20): |
| 238 | """检查图片是否有黑色边框""" |
| 239 | img = Image.open(image_path) |
| 240 | pixels = img.load() |
| 241 | width, height = img.size |
| 242 | |
| 243 | def is_black_pixel(pixel): |
| 244 | return all(x <= black_limit for x in pixel) |
| 245 | |
| 246 | # 检查顶部和底部边框 |
| 247 | for y in range(threshold): |
| 248 | if all(is_black_pixel(pixels[x, y]) for x in range(width)): |
| 249 | return True |
| 250 | if all(is_black_pixel(pixels[x, height - 1 - y]) for x in range(width)): |
| 251 | return True |
| 252 | |
| 253 | # 检查左右边框 |
| 254 | for x in range(threshold): |
| 255 | if all(is_black_pixel(pixels[x, y]) for y in range(height)): |
| 256 | return True |
| 257 | if all(is_black_pixel(pixels[width - 1 - x, y]) for y in range(height)): |
| 258 | return True |
| 259 | |
| 260 | return False |
| 261 | |
| 262 | # ===== 图片上传功能 ===== |
| 263 | |
| 264 | def get_upload_policy(self): |
| 265 | """ |
| 266 | 获取文件上传凭证 |
| 267 | |
| 268 | Returns: |
| 269 | policy_data: 包含上传所需凭证的字典 |
| 270 | |
| 271 | Raises: |
| 272 | Exception: 获取上传凭证失败时 |
| 273 | """ |
| 274 | if not self.api_key: |
| 275 | raise RuntimeError("DASHSCOPE_API_KEY 未设置，无法使用图片上传服务") |
| 276 | |
| 277 | headers = { |
| 278 | "Authorization": f"Bearer {self.api_key}", |
| 279 | "Content-Type": "application/json" |
| 280 | } |
| 281 | params = { |
| 282 | "action": "getPolicy", |
| 283 | "model": self.model_name |
| 284 | } |
| 285 | |
| 286 | response = requests.get( |
| 287 | self.UPLOAD_API_URL, |
| 288 | headers=headers, |
| 289 | params=params, |
| 290 | proxies=self._proxies(), |
| 291 | ) |
| 292 | if response.status_code != 200: |
| 293 | raise Exception(f"Failed to get upload policy: {response.text}") |
| 294 | |
| 295 | return response.json()['data'] |
| 296 | |
| 297 | def upload_file_to_oss(self, policy_data: dict, file_path: str) -> str: |
| 298 | """ |
| 299 | 将文件上传到临时存储OSS |
| 300 | |
| 301 | Args: |
| 302 | policy_data: 上传凭证数据 |
| 303 | file_path: 本地文件路径 |
| 304 | |
| 305 | Returns: |
| 306 | oss_url: OSS URL (格式: oss://...) |
| 307 | |
| 308 | Raises: |
| 309 | Exception: 上传失败时 |
| 310 | """ |
| 311 | file_name = Path(file_path).name |
| 312 | # Sanitize filename for upload to avoid issues with spaces/characters |
| 313 | safe_file_name = "".join([c if c.isalnum() or c in ('-','_','.') else '_' for c in file_name]) |
| 314 | |
| 315 | key = f"{policy_data['upload_dir']}/{safe_file_name}" |
| 316 | |
| 317 | with open(file_path, 'rb') as file: |
| 318 | files = { |
| 319 | 'OSSAccessKeyId': (None, policy_data['oss_access_key_id']), |
| 320 | 'Signature': (None, policy_data['signature']), |
| 321 | 'policy': (None, policy_data['policy']), |
| 322 | 'x-oss-object-acl': (None, policy_data['x_oss_object_acl']), |
| 323 | 'x-oss-forbid-overwrite': (None, policy_data['x_oss_forbid_overwrite']), |
| 324 | 'key': (None, key), |
| 325 | 'success_action_status': (None, '200'), |
| 326 | 'file': (safe_file_name, file) |
| 327 | } |
| 328 | |
| 329 | response = requests.post( |
| 330 | policy_data['upload_host'], |
| 331 | files=files, |
| 332 | proxies=self._proxies(), |
| 333 | ) |
| 334 | if response.status_code != 200: |
| 335 | raise Exception(f"Failed to upload file: {response.text}") |
| 336 | |
| 337 | # Construct OSS URL correctly: oss://<bucket>/<key> |
| 338 | # Extract bucket from upload_host (e.g., https://dashscope-instant.oss-cn-beijing.aliyuncs.com) |
| 339 | upload_host = policy_data['upload_host'] |
| 340 | bucket_name = "" |
| 341 | if '://' in upload_host: |
| 342 | domain = upload_host.split('://')[1] |
| 343 | bucket_name = domain.split('.')[0] |
| 344 | |
| 345 | if bucket_name: |
| 346 | return f"oss://{bucket_name}/{key}" |
| 347 | else: |
| 348 | # Fallback if parsing fails (though unlikely for standard OSS hosts) |
| 349 | # If the original code's assumption that key was self-sufficient was somehow valid, logic is here. |
| 350 | # But normally, oss://<key> is wrong if key doesn't have bucket. |
| 351 | return f"oss://{key}" |
| 352 | |
| 353 | def upload(self, file_path: str) -> str: |
| 354 | """ |
| 355 | 上传文件到阿里云OSS并获取URL（统一接口方法） |
| 356 | |
| 357 | Args: |
| 358 | file_path: 本地文件路径 |
| 359 | |
| 360 | Returns: |
| 361 | oss_url: OSS URL，可在48小时内使用 |
| 362 | |
| 363 | Raises: |
| 364 | FileNotFoundError: 文件不存在时 |
| 365 | RuntimeError: API Key未设置时 |
| 366 | Exception: 上传失败时 |
| 367 | """ |
| 368 | # 检查文件是否存在 |
| 369 | if not os.path.exists(file_path): |
| 370 | raise FileNotFoundError(f"文件不存在: {file_path}") |
| 371 | |
| 372 | if not self.api_key: |
| 373 | raise RuntimeError("DASHSCOPE_API_KEY 未设置，无法使用图片上传服务") |
| 374 | |
| 375 | # 1. 获取上传凭证（注意：上传凭证接口有限流） |
| 376 | policy_data = self.get_upload_policy() |
| 377 | |
| 378 | # 2. 上传文件到OSS |
| 379 | oss_url = self.upload_file_to_oss(policy_data, file_path) |
| 380 | |
| 381 | # 3. 计算过期时间 |
| 382 | expire_time = datetime.now() + timedelta(hours=48) |
| 383 | |
| 384 | logging.info(f"文件上传成功: {file_path}") |
| 385 | logging.info(f" OSS URL: {oss_url}") |
| 386 | logging.info(f" 过期时间: {expire_time.strftime('%Y-%m-%d %H:%M:%S')} (48小时)") |
| 387 | |
| 388 | return oss_url |
| 389 | |
| 390 | def collage_images(self, image_paths, output_path): |
| 391 | """ |
| 392 | 拼图功能：将多张图片水平拼接 |
| 393 | Args: |
| 394 | image_paths: 图片路径列表 |
| 395 | output_path: 输出文件路径 |
| 396 | """ |
| 397 | if not image_paths: |
| 398 | return None |
| 399 | |
| 400 | images = [] |
| 401 | for p in image_paths: |
| 402 | try: |
| 403 | img = Image.open(p) |
| 404 | images.append(img) |
| 405 | except Exception as e: |
| 406 | logging.error(f"Cannot open image {p}: {e}") |
| 407 | |
| 408 | if not images: |
| 409 | return None |
| 410 | |
| 411 | # 统一高度，按第一张图片的高度调整其他图片 |
| 412 | base_height = images[0].height |
| 413 | resized_images = [] |
| 414 | for img in images: |
| 415 | if img.height != base_height: |
| 416 | ratio = base_height / img.height |
| 417 | new_width = int(img.width * ratio) |
| 418 | resized_images.append(img.resize((new_width, base_height))) |
| 419 | else: |
| 420 | resized_images.append(img) |
| 421 | |
| 422 | total_width = sum(img.width for img in resized_images) |
| 423 | new_im = Image.new('RGB', (total_width, base_height)) |
| 424 | |
| 425 | x_offset = 0 |
| 426 | for img in resized_images: |
| 427 | new_im.paste(img, (x_offset, 0)) |
| 428 | x_offset += img.width |
| 429 | |
| 430 | new_im.save(output_path) |
| 431 | return output_path |
| 432 |