| 1 | import os |
| 2 | import sys |
| 3 | |
| 4 | models_dir = os.path.dirname(os.path.abspath(__file__)) |
| 5 | backend_dir = os.path.dirname(models_dir) |
| 6 | if backend_dir not in sys.path: |
| 7 | sys.path.insert(0, backend_dir) |
| 8 | |
| 9 | import requests |
| 10 | import numpy as np |
| 11 | from pathlib import Path |
| 12 | from datetime import datetime, timedelta |
| 13 | from PIL import Image |
| 14 | import logging |
| 15 | from config import Config |
| 16 | |
| 17 | logger = logging.getLogger(__name__) |
| 18 | |
| 19 | |
| 20 | class ImageProcessor: |
| 21 | """ |
| 22 | 图片处理和上传集合类 |
| 23 | 支持：图片处理、分割、拼接，以及上传到阿里云OSS |
| 24 | """ |
| 25 | |
| 26 | # 阿里云DashScope上传配置 |
| 27 | UPLOAD_API_URL = "https://dashscope.aliyuncs.com/api/v1/uploads" |
| 28 | |
| 29 | def __init__(self, |
| 30 | image_path='', |
| 31 | api_key: str = "", |
| 32 | model_name: str = "wan2.6-i2v-flash"): |
| 33 | """ |
| 34 | 初始化图片处理器 |
| 35 | |
| 36 | Args: |
| 37 | image_path: 图片文件路径（可选，用于处理已有图片） |
| 38 | api_key: DashScope API Key（用于上传，可从 config.yaml 读取） |
| 39 | model_name: 模型名称，默认使用 wan2.6-i2v-flash |
| 40 | """ |
| 41 | # 图片处理部分 |
| 42 | if image_path != '': |
| 43 | self.image_path = image_path |
| 44 | self.image = Image.open(image_path) |
| 45 | self.image_np = np.array(self.image) |
| 46 | self.width, self.height = self.image_np.shape[1], self.image_np.shape[0] |
| 47 | else: |
| 48 | self.image_path = None |
| 49 | self.image = None |
| 50 | self.image_np = None |
| 51 | self.width = None |
| 52 | self.height = None |
| 53 | |
| 54 | # 上传功能部分 |
| 55 | self.api_key = api_key or Config.DASHSCOPE_API_KEY |
| 56 | self.model_name = model_name |
| 57 | |
| 58 | @staticmethod |
| 59 | def check_column_white(column_pixels): |
| 60 | """检查列是否几乎全白""" |
| 61 | is_almost_white = np.logical_or(column_pixels == 254, column_pixels == 255) |
| 62 | white_pixels_ratio = np.mean(np.all(is_almost_white, axis=-1)) |
| 63 | return white_pixels_ratio >= 0.98 # 至少98%的像素为白色 |
| 64 | |
| 65 | def find_white_section(self, start, end): |
| 66 | """查找指定范围内的白色区间""" |
| 67 | white_sections = [] |
| 68 | in_white_section = False |
| 69 | start_index = 0 |
| 70 | |
| 71 | for col in range(start, end): |
| 72 | column_pixels = self.image_np[:, col, :] |
| 73 | if self.check_column_white(column_pixels): |
| 74 | if not in_white_section: |
| 75 | start_index = col |
| 76 | in_white_section = True |
| 77 | else: |
| 78 | if in_white_section: |
| 79 | white_sections.append((start_index, col)) |
| 80 | in_white_section = False |
| 81 | |
| 82 | if in_white_section: |
| 83 | white_sections.append((start_index, end)) |
| 84 | |
| 85 | return white_sections |
| 86 | |
| 87 | def split_image(self): |
| 88 | """将图片从中间分割为左右两部分""" |
| 89 | start_col = self.width * 2 // 5 |
| 90 | end_col = self.width * 3 // 5 |
| 91 | white_sections = self.find_white_section(start_col, end_col) |
| 92 | |
| 93 | if white_sections: |
| 94 | middle_section = white_sections[len(white_sections) // 2] |
| 95 | mid_col = (middle_section[0] + middle_section[1]) // 2 |
| 96 | else: |
| 97 | raise ValueError("No suitable white column found within the specified range") |
| 98 | |
| 99 | left_box = (0, 0, mid_col, self.height) |
| 100 | right_box = (mid_col, 0, self.width, self.height) |
| 101 | left_image = self.image.crop(left_box) |
| 102 | right_image = self.image.crop(right_box) |
| 103 | |
| 104 | save_dir, filename = os.path.split(self.image_path) |
| 105 | base, extension = os.path.splitext(filename) |
| 106 | |
| 107 | left_image_path = os.path.join(save_dir, base + '_front' + extension) |
| 108 | right_image_path = os.path.join(save_dir, base + '_back' + extension) |
| 109 | left_image.save(left_image_path) |
| 110 | right_image.save(right_image_path) |
| 111 | |
| 112 | return left_image_path, right_image_path |
| 113 | |
| 114 | def stitch_images(self, image_paths, output_path): |
| 115 | """拼接多张图片""" |
| 116 | if not image_paths: |
| 117 | raise ValueError("No image paths provided") |
| 118 | sample_image = Image.open(image_paths[0]) |
| 119 | single_width, single_height = sample_image.size |
| 120 | num_images = len(image_paths) |
| 121 | total_desired_width = single_width |
| 122 | total_current_width = single_width * num_images |
| 123 | total_width_to_cut = max(0, total_current_width - total_desired_width) |
| 124 | width_to_cut_per_image = total_width_to_cut // num_images |
| 125 | stitched_image = Image.new('RGB', (total_desired_width, single_height), "white") |
| 126 | current_x = 0 |
| 127 | |
| 128 | for path in image_paths: |
| 129 | image = Image.open(path) |
| 130 | if width_to_cut_per_image > 0: |
| 131 | left_margin = width_to_cut_per_image // 2 |
| 132 | right_margin = image.width - width_to_cut_per_image + left_margin |
| 133 | image = image.crop((left_margin, 0, right_margin, image.height)) |
| 134 | stitched_image.paste(image, (current_x, 0)) |
| 135 | current_x += image.width |
| 136 | |
| 137 | output_dir = os.path.dirname(output_path) |
| 138 | if not os.path.exists(output_dir): |
| 139 | os.makedirs(output_dir) |
| 140 | stitched_image.save(output_path) |
| 141 | return output_path |
| 142 | |
| 143 | def download_image(self, image_url, save_path, max_retries=3, proxies=None): |
| 144 | """ |
| 145 | 下载图片，带有重试机制和SSL错误处理 |
| 146 | |
| 147 | Args: |
| 148 | image_url: 图片URL |
| 149 | save_path: 本地保存路径 |
| 150 | max_retries: 最大重试次数 |
| 151 | """ |
| 152 | import time |
| 153 | import urllib3 |
| 154 | |
| 155 | urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning) |
| 156 | |
| 157 | for attempt in range(max_retries): |
| 158 | try: |
| 159 | response = requests.get( |
| 160 | image_url, |
| 161 | timeout=(10, 30), |
| 162 | stream=True, |
| 163 | verify=True, |
| 164 | headers={ |
| 165 | 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36' |
| 166 | }, |
| 167 | proxies=proxies, |
| 168 | ) |
| 169 | |
| 170 | if response.status_code == 200: |
| 171 | with open(save_path, 'wb') as file: |
| 172 | for chunk in response.iter_content(chunk_size=8192): |
| 173 | if chunk: |
| 174 | file.write(chunk) |
| 175 | logger.info("Image downloaded: %s", save_path) |
| 176 | return True |
| 177 | else: |
| 178 | logger.warning("Image download failed: status=%s url=%s", response.status_code, image_url) |
| 179 | |
| 180 | except requests.exceptions.SSLError as e: |
| 181 | logger.warning("Image download SSL error: attempt=%d/%d error=%s", attempt + 1, max_retries, str(e)[:100]) |
| 182 | if attempt < max_retries - 1: |
| 183 | wait_time = (attempt + 1) * 2 |
| 184 | logger.info("Retrying image download in %s seconds", wait_time) |
| 185 | time.sleep(wait_time) |
| 186 | else: |
| 187 | logger.warning("Retrying image download with SSL verification disabled") |
| 188 | try: |
| 189 | response = requests.get( |
| 190 | image_url, |
| 191 | timeout=(10, 30), |
| 192 | stream=True, |
| 193 | verify=False, |
| 194 | headers={ |
| 195 | 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36' |
| 196 | }, |
| 197 | proxies=proxies, |
| 198 | ) |
| 199 | if response.status_code == 200: |
| 200 | with open(save_path, 'wb') as file: |
| 201 | for chunk in response.iter_content(chunk_size=8192): |
| 202 | if chunk: |
| 203 | file.write(chunk) |
| 204 | logger.info("Image downloaded with SSL verification disabled: %s", save_path) |
| 205 | return True |
| 206 | except Exception as fallback_error: |
| 207 | logger.exception("Image download failed after disabling SSL verification: %s", fallback_error) |
| 208 | raise |
| 209 | |
| 210 | except requests.exceptions.Timeout as e: |
| 211 | logger.warning("Image download timeout: attempt=%d/%d error=%s", attempt + 1, max_retries, e) |
| 212 | if attempt < max_retries - 1: |
| 213 | time.sleep((attempt + 1) * 2) |
| 214 | else: |
| 215 | raise |
| 216 | |
| 217 | except Exception as e: |
| 218 | logger.warning("Image download error: attempt=%d/%d error=%s", attempt + 1, max_retries, e) |
| 219 | if attempt < max_retries - 1: |
| 220 | time.sleep((attempt + 1) * 2) |
| 221 | else: |
| 222 | raise |
| 223 | |
| 224 | return False |
| 225 | |
| 226 | def resize_image(self, image_path): |
| 227 | """调整图片大小（添加顶部空白）""" |
| 228 | original_image = Image.open(image_path) |
| 229 | width, height = original_image.size |
| 230 | top_blank_height = height // 2 |
| 231 | final_height = height + top_blank_height |
| 232 | final_width = int(final_height * 5 / 3) |
| 233 | new_image = Image.new("RGB", (final_width, final_height), color="white") |
| 234 | left = (final_width - width) // 2 |
| 235 | top = top_blank_height |
| 236 | new_image.paste(original_image, (left, top)) |
| 237 | new_image.save(image_path) |
| 238 | return image_path |
| 239 | |
| 240 | def has_black_borders(self, image_path, threshold=10, black_limit=20): |
| 241 | """检查图片是否有黑色边框""" |
| 242 | img = Image.open(image_path) |
| 243 | pixels = img.load() |
| 244 | width, height = img.size |
| 245 | |
| 246 | def is_black_pixel(pixel): |
| 247 | return all(x <= black_limit for x in pixel) |
| 248 | |
| 249 | # 检查顶部和底部边框 |
| 250 | for y in range(threshold): |
| 251 | if all(is_black_pixel(pixels[x, y]) for x in range(width)): |
| 252 | return True |
| 253 | if all(is_black_pixel(pixels[x, height - 1 - y]) for x in range(width)): |
| 254 | return True |
| 255 | |
| 256 | # 检查左右边框 |
| 257 | for x in range(threshold): |
| 258 | if all(is_black_pixel(pixels[x, y]) for y in range(height)): |
| 259 | return True |
| 260 | if all(is_black_pixel(pixels[width - 1 - x, y]) for y in range(height)): |
| 261 | return True |
| 262 | |
| 263 | return False |
| 264 | |
| 265 | # ===== 图片上传功能 ===== |
| 266 | |
| 267 | def get_upload_policy(self): |
| 268 | """ |
| 269 | 获取文件上传凭证 |
| 270 | |
| 271 | Returns: |
| 272 | policy_data: 包含上传所需凭证的字典 |
| 273 | |
| 274 | Raises: |
| 275 | Exception: 获取上传凭证失败时 |
| 276 | """ |
| 277 | if not self.api_key: |
| 278 | raise RuntimeError("DASHSCOPE_API_KEY 未设置，无法使用图片上传服务") |
| 279 | |
| 280 | headers = { |
| 281 | "Authorization": f"Bearer {self.api_key}", |
| 282 | "Content-Type": "application/json" |
| 283 | } |
| 284 | params = { |
| 285 | "action": "getPolicy", |
| 286 | "model": self.model_name |
| 287 | } |
| 288 | |
| 289 | response = requests.get( |
| 290 | self.UPLOAD_API_URL, |
| 291 | headers=headers, |
| 292 | params=params, |
| 293 | proxies=Config.requests_proxies("dashscope"), |
| 294 | ) |
| 295 | if response.status_code != 200: |
| 296 | raise Exception(f"Failed to get upload policy: {response.text}") |
| 297 | |
| 298 | return response.json()['data'] |
| 299 | |
| 300 | def upload_file_to_oss(self, policy_data: dict, file_path: str) -> str: |
| 301 | """ |
| 302 | 将文件上传到临时存储OSS |
| 303 | |
| 304 | Args: |
| 305 | policy_data: 上传凭证数据 |
| 306 | file_path: 本地文件路径 |
| 307 | |
| 308 | Returns: |
| 309 | oss_url: OSS URL (格式: oss://...) |
| 310 | |
| 311 | Raises: |
| 312 | Exception: 上传失败时 |
| 313 | """ |
| 314 | file_name = Path(file_path).name |
| 315 | # Sanitize filename for upload to avoid issues with spaces/characters |
| 316 | safe_file_name = "".join([c if c.isalnum() or c in ('-','_','.') else '_' for c in file_name]) |
| 317 | |
| 318 | key = f"{policy_data['upload_dir']}/{safe_file_name}" |
| 319 | |
| 320 | with open(file_path, 'rb') as file: |
| 321 | files = { |
| 322 | 'OSSAccessKeyId': (None, policy_data['oss_access_key_id']), |
| 323 | 'Signature': (None, policy_data['signature']), |
| 324 | 'policy': (None, policy_data['policy']), |
| 325 | 'x-oss-object-acl': (None, policy_data['x_oss_object_acl']), |
| 326 | 'x-oss-forbid-overwrite': (None, policy_data['x_oss_forbid_overwrite']), |
| 327 | 'key': (None, key), |
| 328 | 'success_action_status': (None, '200'), |
| 329 | 'file': (safe_file_name, file) |
| 330 | } |
| 331 | |
| 332 | response = requests.post( |
| 333 | policy_data['upload_host'], |
| 334 | files=files, |
| 335 | proxies=Config.requests_proxies("dashscope"), |
| 336 | ) |
| 337 | if response.status_code != 200: |
| 338 | raise Exception(f"Failed to upload file: {response.text}") |
| 339 | |
| 340 | # Construct OSS URL correctly: oss://<bucket>/<key> |
| 341 | # Extract bucket from upload_host (e.g., https://dashscope-instant.oss-cn-beijing.aliyuncs.com) |
| 342 | upload_host = policy_data['upload_host'] |
| 343 | bucket_name = "" |
| 344 | if '://' in upload_host: |
| 345 | domain = upload_host.split('://')[1] |
| 346 | bucket_name = domain.split('.')[0] |
| 347 | |
| 348 | if bucket_name: |
| 349 | return f"oss://{bucket_name}/{key}" |
| 350 | else: |
| 351 | # Fallback if parsing fails (though unlikely for standard OSS hosts) |
| 352 | # If the original code's assumption that key was self-sufficient was somehow valid, logic is here. |
| 353 | # But normally, oss://<key> is wrong if key doesn't have bucket. |
| 354 | return f"oss://{key}" |
| 355 | |
| 356 | def upload(self, file_path: str) -> str: |
| 357 | """ |
| 358 | 上传文件到阿里云OSS并获取URL（统一接口方法） |
| 359 | |
| 360 | Args: |
| 361 | file_path: 本地文件路径 |
| 362 | |
| 363 | Returns: |
| 364 | oss_url: OSS URL，可在48小时内使用 |
| 365 | |
| 366 | Raises: |
| 367 | FileNotFoundError: 文件不存在时 |
| 368 | RuntimeError: API Key未设置时 |
| 369 | Exception: 上传失败时 |
| 370 | """ |
| 371 | # 检查文件是否存在 |
| 372 | if not os.path.exists(file_path): |
| 373 | raise FileNotFoundError(f"文件不存在: {file_path}") |
| 374 | |
| 375 | if not self.api_key: |
| 376 | raise RuntimeError("DASHSCOPE_API_KEY 未设置，无法使用图片上传服务") |
| 377 | |
| 378 | # 1. 获取上传凭证（注意：上传凭证接口有限流） |
| 379 | policy_data = self.get_upload_policy() |
| 380 | |
| 381 | # 2. 上传文件到OSS |
| 382 | oss_url = self.upload_file_to_oss(policy_data, file_path) |
| 383 | |
| 384 | # 3. 计算过期时间 |
| 385 | expire_time = datetime.now() + timedelta(hours=48) |
| 386 | |
| 387 | logger.info( |
| 388 | "File uploaded to OSS: file=%s oss_url=%s expires_at=%s", |
| 389 | file_path, |
| 390 | oss_url, |
| 391 | expire_time.strftime('%Y-%m-%d %H:%M:%S'), |
| 392 | ) |
| 393 | |
| 394 | return oss_url |
| 395 | |
| 396 | def collage_images(self, image_paths, output_path): |
| 397 | """ |
| 398 | 拼图功能：将多张图片水平拼接 |
| 399 | Args: |
| 400 | image_paths: 图片路径列表 |
| 401 | output_path: 输出文件路径 |
| 402 | """ |
| 403 | if not image_paths: |
| 404 | return None |
| 405 | |
| 406 | images = [] |
| 407 | for p in image_paths: |
| 408 | try: |
| 409 | img = Image.open(p) |
| 410 | images.append(img) |
| 411 | except Exception as e: |
| 412 | logger.error("Cannot open image %s: %s", p, e) |
| 413 | |
| 414 | if not images: |
| 415 | return None |
| 416 | |
| 417 | # 统一高度，按第一张图片的高度调整其他图片 |
| 418 | base_height = images[0].height |
| 419 | resized_images = [] |
| 420 | for img in images: |
| 421 | if img.height != base_height: |
| 422 | ratio = base_height / img.height |
| 423 | new_width = int(img.width * ratio) |
| 424 | resized_images.append(img.resize((new_width, base_height))) |
| 425 | else: |
| 426 | resized_images.append(img) |
| 427 | |
| 428 | total_width = sum(img.width for img in resized_images) |
| 429 | new_im = Image.new('RGB', (total_width, base_height)) |
| 430 | |
| 431 | x_offset = 0 |
| 432 | for img in resized_images: |
| 433 | new_im.paste(img, (x_offset, 0)) |
| 434 | x_offset += img.width |
| 435 | |
| 436 | new_im.save(output_path) |
| 437 | return output_path |
| 438 |