Initial commit

.gitignore

+# Python
+backend/.venv/
+backend/.venv310/
+backend/__pycache__/
+backend/**/*.pyc
+backend/.pytest_cache/
+
+# Node
+frontend/node_modules/
+frontend/dist/
+frontend/.vite/
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+
+# OS
+.DS_Store
+Thumbs.db
+
+# Temporary files
+*.tmp
+*.log

README.md

+# 文档去水印工具
+
+一个基于 Vue 3 + Flask 的本地文档去水印网站，支持自动识别并删除 DOC、DOCX、PDF 文件中的文字与图片水印。
+
+## 功能特性
+
+- 支持格式：`doc`、`docx`、`pdf`（扩展名大小写不敏感）
+- 全自动水印识别，无需用户输入水印内容
+- 支持文字水印与图片水印
+- DOC 文件通过 LibreOffice 转换为 DOCX 处理后转回 DOC
+- 处理完成后立即清理临时文件，服务器不保留任何上传文件或结果文件
+- 结果文件名自动加 `_` 前缀，例如 `报告.docx` → `_报告.docx`
+- 上传文件大小限制：20MB
+
+## 技术栈
+
+- 前端：Vue 3 + Vite
+- 后端：Python 3.8+ + Flask
+- 依赖库：
+  - `pikepdf` / `pdfplumber`：PDF 处理
+  - `lxml` / `python-docx`：DOCX 处理
+  - `Flask-CORS`：跨域支持
+
+## 目录结构
+
+```
+WaterMarkRemoveTool/
+├── backend/                  # Flask 后端
+│   ├── app.py                # Flask 应用入口
+│   ├── config.py             # 配置（格式、大小、关键词等）
+│   ├── requirements.txt      # Python 依赖
+│   ├── handlers/             # 各格式处理器
+│   │   ├── docx_handler.py   # DOCX 去水印
+│   │   ├── doc_handler.py    # DOC 去水印（调用 LibreOffice）
+│   │   └── pdf_handler.py    # PDF 去水印
+│   └── utils/                # 工具函数
+│       ├── file_utils.py     # 临时目录、命名、清理
+│       └── watermark_detector.py  # 水印特征识别
+└── frontend/                 # Vue 3 前端
+    ├── index.html
+    ├── package.json
+    ├── vite.config.js        # 开发代理配置
+    └── src/
+        ├── App.vue
+        ├── main.js
+        ├── api/watermark.js  # 后端 API 调用
+        └── components/FileUploader.vue
+```
+
+## 环境要求
+
+- Python 3.8+
+- Node.js 18+
+- LibreOffice（用于 DOC 转换，安装后确保 `soffice` 命令可用）
+
+### 验证 LibreOffice
+
+```bash
+soffice --version
+```
+
+## 安装与运行
+
+### 1. 克隆/进入项目目录
+
+```bash
+cd /home/hunter/Desktop/WaterMarkRemoveTool
+```
+
+### 2. 启动后端
+
+```bash
+cd backend
+python3 -m venv venv
+source venv/bin/activate
+pip install -r requirements.txt
+python app.py
+```
+
+后端默认运行在 `http://127.0.0.1:5000`。
+
+### 3. 启动前端（开发环境）
+
+在另一个终端中：
+
+```bash
+cd frontend
+npm install
+npm run dev
+```
+
+前端默认运行在 `http://127.0.0.1:5173`，并通过 Vite 代理将 `/api` 请求转发到后端。
+
+打开浏览器访问 `http://127.0.0.1:5173` 即可使用。
+
+## 生产部署
+
+```bash
+cd frontend
+npm run build
+
+cd ../backend
+source venv/bin/activate
+python app.py
+```
+
+生产模式下，Flask 会直接提供 `frontend/dist` 目录中的静态文件。
+
+## 主要 API
+
+### `POST /api/remove-watermark`
+
+上传文件并去除水印。
+
+- Content-Type: `multipart/form-data`
+- Field: `file`
+- 成功：返回 `200` + 处理后的文件二进制流
+  - 响应头：`Content-Disposition: attachment; filename="_原始文件名"`
+- 失败：返回 `400` JSON
+  - `{ "message": "错误信息" }`
+
+### `GET /api/health`
+
+健康检查。
+
+- 返回：`{ "status": "ok" }`
+
+## 去水印策略
+
+### DOCX
+
+1. 使用 `zipfile` 解压 DOCX 文件。
+2. 解析 `word/header*.xml` 和 `word/footer*.xml`。
+3. 删除包含水印关键词的段落。
+4. 删除 VML/DrawingML 水印 shape。
+5. 删除孤立的图片水印段落。
+6. 清理被引用但已删除的媒体文件及 relationships。
+7. 重新打包为 DOCX。
+
+### DOC
+
+1. 使用 LibreOffice 将 DOC 转换为 DOCX。
+2. 对 DOCX 执行上述去水印逻辑。
+3. 再次使用 LibreOffice 将处理后的 DOCX 转回 DOC。
+
+> 注意：DOC 转换过程可能轻微改变排版、字体等内容。
+
+### PDF
+
+1. 删除 `/Subtype /Watermark` 的 annotation。
+2. 关闭或删除 Optional Content（图层）水印。
+3. 使用 `pdfplumber` 识别文本水印，并在内容流中清除对应文本。
+4. 删除在每一页重复出现或带有透明掩膜的图片 XObject。
+
+## 配置说明
+
+可在 `backend/config.py` 中修改：
+
+- `ALLOWED_EXTENSIONS`：允许上传的文件扩展名
+- `MAX_CONTENT_LENGTH`：最大上传文件大小（默认 20MB）
+- `SOFFICE_TIMEOUT`：LibreOffice 转换超时时间（默认 120 秒）
+- `WATERMARK_KEYWORDS`：用于启发式识别水印的关键词列表
+
+## 限制与说明
+
+- 不保证 100% 去除任意水印，尤其是复杂 PDF 水印或自定义图片水印。
+- 全自动图片水印识别基于启发式规则，可能误删正文图片或漏删水印。
+- 直接修改 PDF 内容流可能导致某些复杂 PDF 显示异常。
+- DOC 文件经 LibreOffice 转换后排版可能存在细微变化。
+- 不做用户系统、历史记录、文件持久化。
+
+## 常见问题
+
+### 上传 DOC 文件提示 LibreOffice 未安装
+
+请确保系统已安装 LibreOffice：
+
+```bash
+sudo apt update
+sudo apt install libreoffice
+```
+
+安装后验证：
+
+```bash
+soffice --version
+```
+
+### 前端无法连接后端
+
+开发模式下，Vite 已配置代理将 `/api` 转发到 `http://127.0.0.1:5000`。请确保后端已启动。
+
+### 处理后的 PDF 显示异常
+
+某些复杂 PDF 的内容流结构特殊，自动清除水印文本时可能误伤正文。建议先用简单 PDF 测试。
+
+## 开发与调试
+
+```bash
+# 后端调试
+cd backend
+source venv/bin/activate
+python app.py
+
+# 前端调试
+cd frontend
+npm run dev
+```
+
+## License
+
+本项目仅供学习和本地使用。

backend/app.py

+import os
+import shutil
+import traceback
+from pathlib import Path
+
+from flask import Flask, after_this_request, jsonify, request, send_file, send_from_directory
+from flask_cors import CORS
+from werkzeug.utils import secure_filename
+
+from config import ALLOWED_EXTENSIONS, MAX_CONTENT_LENGTH
+from handlers.doc_handler import remove_watermark as remove_doc_watermark
+from handlers.docx_handler import remove_watermark as remove_docx_watermark
+from handlers.pdf_handler import remove_watermark as remove_pdf_watermark
+from utils.file_utils import (
+    cleanup_dir,
+    is_allowed_file,
+    make_temp_dir,
+    prefixed_output_name,
+)
+
+BASE_DIR = Path(__file__).resolve().parent
+FRONTEND_DIST = BASE_DIR.parent / "frontend" / "dist"
+
+app = Flask(__name__, static_folder=str(FRONTEND_DIST) if FRONTEND_DIST.exists() else None)
+app.config["MAX_CONTENT_LENGTH"] = MAX_CONTENT_LENGTH
+CORS(app, resources={r"/api/*": {"origins": "*"}})
+
+
+@app.errorhandler(413)
+def request_entity_too_large(error):
+    return jsonify({"message": "文件大小超过限制（最大 20MB）"}), 413
+
+
+@app.route("/api/remove-watermark", methods=["POST"])
+def remove_watermark():
+    temp_dir = None
+    try:
+        if "file" not in request.files:
+            return jsonify({"message": "未找到上传文件"}), 400
+
+        file = request.files["file"]
+        if not file or not file.filename:
+            return jsonify({"message": "未选择文件"}), 400
+
+        original_filename = file.filename
+        if not is_allowed_file(original_filename, ALLOWED_EXTENSIONS):
+            return jsonify({"message": "仅支持 doc、docx、pdf 格式"}), 400
+
+        temp_dir = make_temp_dir()
+        ext = original_filename.rsplit(".", 1)[1].lower()
+        safe_name = secure_filename(original_filename)
+        input_path = temp_dir / safe_name
+        file.save(str(input_path))
+
+        output_name = prefixed_output_name(original_filename)
+        output_path = temp_dir / secure_filename(output_name)
+
+        if ext == "docx":
+            remove_docx_watermark(input_path, output_path)
+        elif ext == "doc":
+            remove_doc_watermark(input_path, output_path)
+        elif ext == "pdf":
+            remove_pdf_watermark(input_path, output_path)
+        else:
+            return jsonify({"message": "不支持的文件格式"}), 400
+
+        if not output_path.exists():
+            return jsonify({"message": "处理失败，未生成输出文件"}), 500
+
+        @after_this_request
+        def cleanup(response):
+            if temp_dir and temp_dir.exists():
+                shutil.rmtree(temp_dir, ignore_errors=True)
+            return response
+
+        return send_file(
+            str(output_path),
+            as_attachment=True,
+            download_name=output_name,
+            mimetype="application/octet-stream",
+        )
+    except RuntimeError as exc:
+        if temp_dir:
+            cleanup_dir(temp_dir)
+        return jsonify({"message": str(exc)}), 500
+    except Exception as exc:
+        if temp_dir:
+            cleanup_dir(temp_dir)
+        traceback.print_exc()
+        return jsonify({"message": f"处理失败: {str(exc)}"}), 500
+
+
+@app.route("/api/health", methods=["GET"])
+def health():
+    return jsonify({"status": "ok"}), 200
+
+
+@app.route("/", defaults={"path": ""})
+@app.route("/<path:path>")
+def serve_frontend(path):
+    """生产环境：提供前端静态文件。"""
+    if not FRONTEND_DIST.exists():
+        return jsonify({"message": "前端资源未构建"}), 404
+    target = FRONTEND_DIST / path
+    if path and target.exists() and target.is_file():
+        return send_from_directory(FRONTEND_DIST, path)
+    return send_from_directory(FRONTEND_DIST, "index.html")
+
+
+if __name__ == "__main__":
+    # 开发环境默认端口 5000
+    port = int(os.environ.get("PORT", 5000))
+    app.run(host="0.0.0.0", port=port, debug=True)

backend/config.py

+import os
+from pathlib import Path
+
+BASE_DIR = Path(__file__).resolve().parent
+
+# 允许格式（扩展名大小写不敏感）
+ALLOWED_EXTENSIONS = {"doc", "docx", "pdf"}
+
+# 上传文件大小限制（字节）
+MAX_CONTENT_LENGTH = 20 * 1024 * 1024  # 20 MB
+
+# LibreOffice 转换超时（秒）
+SOFFICE_TIMEOUT = 120
+
+# 常见水印关键词（用于启发式识别）
+WATERMARK_KEYWORDS = {
+    "机密", "秘密", "绝密", "内部", "样稿", "样张", "草稿", "草案",
+    "draft", "confidential", "sample", "prototype", "internal",
+    " watermark", "watermark", "do not copy", "top secret",
+}
+
+# 水印常见透明度阈值
+WATERMARK_ALPHA_THRESHOLD = 0.6
+
+# 重复 Form XObject 水印检测阈值
+REPEATED_FORM_MIN_PAGES = 2
+REPEATED_FORM_MIN_COVERAGE = 0.90
+REPEATED_FORM_MAX_BBOX_AREA_RATIO = 0.25
+REPEATED_FORM_SMALL_BBOX_AREA_RATIO = 0.05
+REPEATED_FORM_WATERMARK_SCORE_THRESHOLD = 3

backend/handlers/doc_handler.py

+import shutil
+import subprocess
+from pathlib import Path
+from typing import List
+
+from config import SOFFICE_TIMEOUT
+from handlers.docx_handler import remove_watermark as remove_docx_watermark
+
+
+def _run_soffice(args: List[str], timeout: int = SOFFICE_TIMEOUT) -> None:
+    """调用 LibreOffice 无头模式。"""
+    cmd = ["soffice", "--headless"] + args
+    try:
+        subprocess.run(
+            cmd,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+            check=True,
+            timeout=timeout,
+        )
+    except FileNotFoundError as exc:
+        raise RuntimeError("LibreOffice 未安装或未找到 soffice 命令") from exc
+    except subprocess.TimeoutExpired as exc:
+        raise RuntimeError("LibreOffice 转换超时") from exc
+    except subprocess.CalledProcessError as exc:
+        err = exc.stderr.decode("utf-8", errors="ignore") if exc.stderr else str(exc)
+        raise RuntimeError(f"LibreOffice 转换失败: {err}") from exc
+
+
+def remove_watermark(input_path: Path, output_path: Path) -> None:
+    """DOC 文件：先转 DOCX，去水印，再转回 DOC。"""
+    temp_dir = input_path.parent
+    docx_path = temp_dir / (input_path.stem + ".docx")
+    cleaned_docx = temp_dir / (input_path.stem + "_cleaned.docx")
+
+    # 1) DOC -> DOCX
+    _run_soffice([
+        "--convert-to", "docx",
+        "--outdir", str(temp_dir),
+        str(input_path),
+    ])
+
+    if not docx_path.exists():
+        raise RuntimeError("DOC 转换为 DOCX 失败")
+
+    # 2) 去水印
+    remove_docx_watermark(docx_path, cleaned_docx)
+
+    # 3) DOCX -> DOC
+    _run_soffice([
+        "--convert-to", "doc",
+        "--outdir", str(temp_dir),
+        str(cleaned_docx),
+    ])
+
+    # LibreOffice 转换后生成的文件名与 cleaned_docx 的 stem 相同
+    generated_doc = temp_dir / (cleaned_docx.stem + ".doc")
+    if not generated_doc.exists():
+        raise RuntimeError("DOCX 转换回 DOC 失败")
+
+    shutil.copy2(generated_doc, output_path)

backend/handlers/docx_handler.py

+import re
+import shutil
+import zipfile
+from pathlib import Path
+from typing import Set
+from xml.etree import ElementTree as ET
+
+from utils.watermark_detector import (
+    NSMAP,
+    gather_image_rels_in_element,
+    is_watermark_shape,
+    looks_like_watermark_text,
+    register_all_namespaces,
+)
+
+
+def _tag(name: str, prefix: str = "w") -> str:
+    return f"{{{NSMAP[prefix]}}}{name}"
+
+
+def _register_namespaces_from_xml(xml_text: str) -> None:
+    """从 XML 文本中提取 xmlns 声明并注册，避免写回时前缀被改写。"""
+    for match in re.finditer(r'xmlns:([a-zA-Z0-9_]+)="([^"]+)"', xml_text):
+        prefix, uri = match.groups()
+        ET.register_namespace(prefix, uri)
+
+
+def remove_watermark(input_path: Path, output_path: Path) -> None:
+    """去除 DOCX 文件中的文字和图片水印。"""
+    register_all_namespaces()
+
+    work_dir = input_path.parent / "docx_extracted"
+    work_dir.mkdir(parents=True, exist_ok=True)
+
+    try:
+        with zipfile.ZipFile(input_path, "r") as zin:
+            zin.extractall(work_dir)
+
+        word_dir = work_dir / "word"
+        rels_dir = word_dir / "_rels"
+        media_dir = word_dir / "media"
+
+        # 处理 header/footer XML
+        target_files = []
+        if word_dir.exists():
+            for f in word_dir.iterdir():
+                if f.is_file() and f.name.lower().startswith(("header", "footer")) and f.suffix == ".xml":
+                    target_files.append(f)
+
+        removed_rels: Set[str] = set()
+
+        for xml_file in target_files:
+            try:
+                xml_text = xml_file.read_text(encoding="utf-8")
+                _register_namespaces_from_xml(xml_text)
+                root = ET.fromstring(xml_text)
+
+                changed = False
+
+                # 1) 删除包含水印文字的段落 / shape
+                # 页眉页脚中的段落
+                for p in root.findall(".//w:p", NSMAP):
+                    text_nodes = p.findall(".//w:t", NSMAP)
+                    text = "".join(t.text or "" for t in text_nodes)
+                    if looks_like_watermark_text(text):
+                        # 收集该段落中引用的图片 rels
+                        removed_rels |= gather_image_rels_in_element(p, NSMAP)
+                        # 删除段落
+                        parent = None
+                        for parent_candidate in root.iter():
+                            if p in list(parent_candidate):
+                                parent = parent_candidate
+                                break
+                        if parent is not None:
+                            parent.remove(p)
+                            changed = True
+
+                # 2) 删除 VML/DrawingML 水印 shape
+                for shape in root.findall(".//v:shape", NSMAP):
+                    if is_watermark_shape(shape, NSMAP):
+                        removed_rels |= gather_image_rels_in_element(shape, NSMAP)
+                        parent = None
+                        for parent_candidate in root.iter():
+                            if shape in list(parent_candidate):
+                                parent = parent_candidate
+                                break
+                        if parent is not None:
+                            parent.remove(shape)
+                            changed = True
+
+                # 3) 删除独立的图片水印（仅出现在 header/footer 中且大面积覆盖的通常难以自动识别，
+                #    这里删除看起来“孤立”的图片：即所在段落除图片外无其他可见文本且图片没有正文字号）
+                for pict in root.findall(".//w:pict", NSMAP):
+                    # 判断其所在段落是否已无文本
+                    parent_p = None
+                    for parent_candidate in root.iter():
+                        if pict in list(parent_candidate):
+                            parent_p = parent_candidate
+                            break
+                    if parent_p is not None:
+                        texts = parent_p.findall(".//w:t", NSMAP)
+                        if all(not (t.text or "").strip() for t in texts):
+                            removed_rels |= gather_image_rels_in_element(pict, NSMAP)
+                            parent_p_parent = None
+                            for ppc in root.iter():
+                                if parent_p in list(ppc):
+                                    parent_p_parent = ppc
+                                    break
+                            if parent_p_parent is not None:
+                                parent_p_parent.remove(parent_p)
+                                changed = True
+
+                if changed:
+                    tree = ET.ElementTree(root)
+                    tree.write(xml_file, encoding="UTF-8", xml_declaration=True)
+            except Exception:
+                # 解析失败时保留原文件，避免整个请求失败
+                pass
+
+        # 4) 删除被引用的 media 文件以及对应的 relationships
+        if removed_rels and rels_dir.exists():
+            rels_file = rels_dir / "document.xml.rels"
+            if not rels_file.exists():
+                rels_file = rels_dir / "document2.xml.rels"
+            # 实际上 header/footer 的 rels 分别存在各自的 .rels 中
+            for rels_f in rels_dir.iterdir():
+                if rels_f.suffix != ".rels":
+                    continue
+                try:
+                    rels_text = rels_f.read_text(encoding="utf-8")
+                    _register_namespaces_from_xml(rels_text)
+                    root = ET.fromstring(rels_text)
+                    rel_ns = "http://schemas.openxmlformats.org/package/2006/relationships"
+                    for rel in root.findall(f"{{{rel_ns}}}Relationship"):
+                        rid = rel.get("Id")
+                        if rid in removed_rels:
+                            target = rel.get("Target")
+                            root.remove(rel)
+                            # 删除 media 文件
+                            if target and media_dir.exists():
+                                target_clean = target
+                                if target_clean.startswith("media/"):
+                                    target_clean = target_clean[6:]
+                                media_path = media_dir / target_clean
+                                if media_path.exists():
+                                    media_path.unlink()
+                    tree = ET.ElementTree(root)
+                    tree.write(rels_f, encoding="UTF-8", xml_declaration=True)
+                except Exception:
+                    pass
+
+        # 重新打包为 DOCX
+        output_path.parent.mkdir(parents=True, exist_ok=True)
+        with zipfile.ZipFile(output_path, "w", zipfile.ZIP_DEFLATED) as zout:
+            for file_path in work_dir.rglob("*"):
+                if file_path.is_file():
+                    arcname = file_path.relative_to(work_dir)
+                    zout.write(file_path, arcname)
+    finally:
+        if work_dir.exists():
+            shutil.rmtree(work_dir, ignore_errors=True)

backend/requirements.txt

+Flask>=3.0
+Flask-CORS>=4.0
+Werkzeug>=3.0
+python-docx>=1.1
+pikepdf>=8.0
+pdfplumber>=0.10
+lxml>=5.0

backend/utils/file_utils.py

+import os
+import shutil
+import tempfile
+from pathlib import Path
+from typing import Optional, Set
+
+from werkzeug.utils import secure_filename
+
+
+def is_allowed_file(filename: str, allowed_extensions: Set[str]) -> bool:
+    """校验文件扩展名是否在允许列表内（大小写不敏感）。"""
+    if not filename or "." not in filename:
+        return False
+    ext = filename.rsplit(".", 1)[1].lower()
+    return ext in allowed_extensions
+
+
+def sanitize_filename(filename: str) -> str:
+    """清理文件名，保留原始扩展名。"""
+    return secure_filename(filename)
+
+
+def prefixed_output_name(filename: str) -> str:
+    """结果文件名：原始文件名前加 '_'。"""
+    name = filename.strip()
+    if not name:
+        return "_output"
+    # 去掉路径
+    name = os.path.basename(name)
+    # 若已有前缀，不再重复添加
+    if name.startswith("_"):
+        return name
+    return "_" + name
+
+
+def make_temp_dir() -> Path:
+    """创建临时目录，返回 Path 对象。"""
+    return Path(tempfile.mkdtemp(prefix="wmr_"))
+
+
+def cleanup_dir(directory: Optional[Path]) -> None:
+    """清理临时目录。"""
+    if directory and directory.exists():
+        shutil.rmtree(directory, ignore_errors=True)

backend/utils/watermark_detector.py

+import re
+from pathlib import Path
+from typing import Dict, Set
+
+from config import WATERMARK_ALPHA_THRESHOLD, WATERMARK_KEYWORDS
+
+NSMAP = {
+    "w": "http://schemas.openxmlformats.org/wordprocessingml/2006/main",
+    "v": "urn:schemas-microsoft-com:vml",
+    "o": "urn:schemas-microsoft-com:office:office",
+    "r": "http://schemas.openxmlformats.org/officeDocument/2006/relationships",
+    "wp": "http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing",
+    "a": "http://schemas.openxmlformats.org/drawingml/2006/main",
+    "pic": "http://schemas.openxmlformats.org/drawingml/2006/picture",
+    "w10": "urn:schemas-microsoft-com:office:word",
+}
+
+
+def register_all_namespaces():
+    """在 lxml 中注册常用命名空间，避免写回时前缀被改写。"""
+    try:
+        from lxml import etree
+        for prefix, uri in NSMAP.items():
+            etree.register_namespace(prefix, uri)
+    except Exception:
+        pass
+
+
+def looks_like_watermark_text(text: str) -> bool:
+    """判断文本是否像水印文字。"""
+    if not text:
+        return False
+    lower = text.lower()
+    for kw in WATERMARK_KEYWORDS:
+        if kw in lower:
+            return True
+    # 某些全角/半角通用的“水印”字样
+    if "水印" in text:
+        return True
+    return False
+
+
+def is_watermark_shape(elem, namespaces: Dict[str, str]) -> bool:
+    """判断 DOCX 中 VML/DrawingML shape 是否为水印。"""
+    try:
+        # VML 文字水印通常 type="#_x0000_t136"
+        shape_type = elem.get("type")
+        if shape_type and "t136" in shape_type:
+            return True
+        # style 中包含 opacity 较低
+        style = elem.get("style") or ""
+        if "opacity" in style.lower():
+            m = re.search(r"opacity[:=\s]*([0-9.]+)", style, re.I)
+            if m and float(m.group(1)) < WATERMARK_ALPHA_THRESHOLD:
+                return True
+        # 子节点包含水印关键词
+        text_nodes = elem.xpath(".//v:textpath | .//w:t | .//v:t", namespaces=namespaces)
+        for t in text_nodes:
+            val = t.text or t.get("string") or ""
+            if looks_like_watermark_text(val):
+                return True
+    except Exception:
+        pass
+    return False
+
+
+def gather_image_rels_in_element(elem, namespaces: Dict[str, str]) -> Set[str]:
+    """收集元素内引用的所有图片 relationship id。"""
+    rels = set()
+    try:
+        r_ns = "{http://schemas.openxmlformats.org/officeDocument/2006/relationships}"
+        for node in elem.xpath(".//a:blip | .//v:imagedata", namespaces=namespaces):
+            # DrawingML 图片通常使用 r:embed
+            rel = node.get(f"{r_ns}embed")
+            if rel:
+                rels.add(rel)
+            # VML 图片通常使用 r:id
+            rel = node.get(f"{r_ns}id")
+            if rel:
+                rels.add(rel)
+            # 兜底：任意 r: 开头的 relationship 属性
+            for attr, val in node.attrib.items():
+                if attr.startswith(r_ns) and val:
+                    rels.add(val)
+    except Exception:
+        pass
+    return rels

frontend/index.html

+<!DOCTYPE html>
+<html lang="zh-CN">
+  <head>
+    <meta charset="UTF-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <title>去水印工具</title>
+  </head>
+  <body>
+    <div id="app"></div>
+    <script type="module" src="/src/main.js"></script>
+  </body>
+</html>

frontend/package.json

+{
+  "name": "watermark-remove-frontend",
+  "private": true,
+  "version": "1.0.0",
+  "type": "module",
+  "scripts": {
+    "dev": "vite",
+    "build": "vite build",
+    "preview": "vite preview"
+  },
+  "dependencies": {
+    "vue": "^3.4.0"
+  },
+  "devDependencies": {
+    "@vitejs/plugin-vue": "^5.0.0",
+    "vite": "^5.0.0"
+  }
+}

frontend/src/App.vue

+<template>
+  <div class="app">
+    <header class="header">
+      <h1>文档去水印工具</h1>
+      <p class="subtitle">全自动识别并删除 DOC、DOCX、PDF 文件中的文字与图片水印</p>
+    </header>
+    <main class="main">
+      <FileUploader />
+      <div class="notes">
+        <h3>使用说明</h3>
+        <ul>
+          <li>上传文件后将由服务器自动处理，处理完成后自动下载。</li>
+          <li>下载文件名将自动加上下划线前缀，例如 <code>报告.docx</code> → <code>_报告.docx</code>。</li>
+          <li>服务器不会保留您的原始文件与结果文件，处理完成后立即删除。</li>
+          <li>对于复杂 PDF 水印或自定义图片水印，去除效果可能有限。</li>
+          <li>DOC 文件会经过 LibreOffice 转换为 DOCX 处理后再转回 DOC，排版可能存在细微变化。</li>
+        </ul>
+      </div>
+    </main>
+    <footer class="footer">
+      <p>本地运行 · 数据不留存</p>
+    </footer>
+  </div>
+</template>
+
+<script setup>
+import FileUploader from './components/FileUploader.vue'
+</script>
+
+<style>
+* {
+  box-sizing: border-box;
+}
+
+body {
+  margin: 0;
+  background-color: #f5f7fa;
+  color: #303133;
+}
+
+.app {
+  min-height: 100vh;
+  display: flex;
+  flex-direction: column;
+}
+
+.header {
+  background-color: #fff;
+  padding: 40px 20px 24px;
+  text-align: center;
+  box-shadow: 0 2px 12px rgba(0, 0, 0, 0.06);
+}
+
+.header h1 {
+  margin: 0 0 8px;
+  font-size: 28px;
+  color: #303133;
+}
+
+.subtitle {
+  margin: 0;
+  color: #606266;
+  font-size: 14px;
+}
+
+.main {
+  flex: 1;
+  padding: 32px 20px;
+}
+
+.notes {
+  max-width: 520px;
+  margin: 32px auto 0;
+  padding: 20px 24px;
+  background-color: #fff;
+  border-radius: 12px;
+  box-shadow: 0 2px 12px rgba(0, 0, 0, 0.06);
+}
+
+.notes h3 {
+  margin: 0 0 12px;
+  font-size: 16px;
+}
+
+.notes ul {
+  margin: 0;
+  padding-left: 20px;
+  font-size: 13px;
+  color: #606266;
+  line-height: 1.8;
+}
+
+.notes code {
+  background-color: #f4f4f5;
+  padding: 2px 6px;
+  border-radius: 4px;
+  font-family: Menlo, Monaco, Consolas, monospace;
+}
+
+.footer {
+  text-align: center;
+  padding: 20px;
+  color: #909399;
+  font-size: 12px;
+}
+</style>

frontend/src/api/watermark.js

+const API_BASE = import.meta.env.VITE_API_BASE || ''
+
+export async function removeWatermark(file, onProgress) {
+  const formData = new FormData()
+  formData.append('file', file)
+
+  const response = await fetch(`${API_BASE}/api/remove-watermark`, {
+    method: 'POST',
+    body: formData,
+  })
+
+  if (!response.ok) {
+    let message = '处理失败'
+    try {
+      const err = await response.json()
+      message = err.message || message
+    } catch {
+      message = `${message} (${response.status})`
+    }
+    throw new Error(message)
+  }
+
+  const blob = await response.blob()
+  const disposition = response.headers.get('Content-Disposition') || ''
+  let filename = file.name
+  const match = disposition.match(/filename[^;=\n]*=((['"]).*?\2|[^;\n]*)/)
+  if (match && match[1]) {
+    filename = decodeURIComponent(match[1].replace(/['"]/g, ''))
+  }
+
+  return { blob, filename }
+}

frontend/src/components/FileUploader.vue

+<template>
+  <div class="uploader">
+    <div
+      class="drop-zone"
+      :class="{ dragging: isDragging, disabled: isProcessing }"
+      @dragenter.prevent="isDragging = true"
+      @dragleave.prevent="isDragging = false"
+      @dragover.prevent
+      @drop.prevent="handleDrop"
+      @click="triggerFileInput"
+    >
+      <input
+        ref="fileInput"
+        type="file"
+        accept=".doc,.docx,.pdf"
+        class="hidden-input"
+        @change="handleFileChange"
+      />
+      <p v-if="!selectedFile">点击或拖拽上传文件</p>
+      <p v-else class="file-name">已选择：{{ selectedFile.name }}</p>
+      <p class="hint">支持格式：doc、docx、pdf（最大 20MB）</p>
+    </div>
+
+    <div class="actions">
+      <button
+        class="btn primary"
+        :disabled="!selectedFile || isProcessing"
+        @click="submit"
+      >
+        {{ isProcessing ? '处理中...' : '去除水印' }}
+      </button>
+      <button
+        class="btn"
+        :disabled="isProcessing"
+        @click="reset"
+      >
+        重置
+      </button>
+    </div>
+
+    <div v-if="message" class="message" :class="messageType">
+      {{ message }}
+    </div>
+
+    <div v-if="isProcessing" class="progress">
+      <div class="progress-bar"></div>
+    </div>
+  </div>
+</template>
+
+<script setup>
+import { ref } from 'vue'
+import { removeWatermark } from '../api/watermark.js'
+
+const ALLOWED_EXTENSIONS = ['doc', 'docx', 'pdf']
+const MAX_SIZE = 20 * 1024 * 1024
+
+const fileInput = ref(null)
+const selectedFile = ref(null)
+const isDragging = ref(false)
+const isProcessing = ref(false)
+const message = ref('')
+const messageType = ref('')
+
+function isAllowed(filename) {
+  const ext = filename.split('.').pop()?.toLowerCase()
+  return ALLOWED_EXTENSIONS.includes(ext)
+}
+
+function setMessage(text, type = 'info') {
+  message.value = text
+  messageType.value = type
+}
+
+function handleFile(file) {
+  if (!file) return
+  if (!isAllowed(file.name)) {
+    setMessage('仅支持 doc、docx、pdf 格式', 'error')
+    selectedFile.value = null
+    return
+  }
+  if (file.size > MAX_SIZE) {
+    setMessage('文件大小超过 20MB 限制', 'error')
+    selectedFile.value = null
+    return
+  }
+  selectedFile.value = file
+  setMessage('')
+}
+
+function handleDrop(e) {
+  isDragging.value = false
+  const file = e.dataTransfer?.files?.[0]
+  handleFile(file)
+}
+
+function handleFileChange(e) {
+  const file = e.target.files?.[0]
+  handleFile(file)
+}
+
+function triggerFileInput() {
+  if (isProcessing.value) return
+  fileInput.value?.click()
+}
+
+function downloadBlob(blob, filename) {
+  const url = URL.createObjectURL(blob)
+  const a = document.createElement('a')
+  a.href = url
+  a.download = filename
+  document.body.appendChild(a)
+  a.click()
+  document.body.removeChild(a)
+  URL.revokeObjectURL(url)
+}
+
+async function submit() {
+  if (!selectedFile.value || isProcessing.value) return
+  isProcessing.value = true
+  setMessage('')
+
+  try {
+    const { blob, filename } = await removeWatermark(selectedFile.value)
+    downloadBlob(blob, filename)
+    setMessage(`处理完成：${filename}`, 'success')
+  } catch (err) {
+    setMessage(err.message || '处理失败', 'error')
+  } finally {
+    isProcessing.value = false
+  }
+}
+
+function reset() {
+  selectedFile.value = null
+  message.value = ''
+  if (fileInput.value) {
+    fileInput.value.value = ''
+  }
+}
+</script>
+
+<style scoped>
+.uploader {
+  max-width: 520px;
+  margin: 0 auto;
+  font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, 'Helvetica Neue', Arial, sans-serif;
+}
+
+.drop-zone {
+  border: 2px dashed #c0c4cc;
+  border-radius: 12px;
+  padding: 48px 24px;
+  text-align: center;
+  cursor: pointer;
+  transition: border-color 0.2s, background-color 0.2s;
+  background-color: #fafafa;
+}
+
+.drop-zone:hover,
+.drop-zone.dragging {
+  border-color: #409eff;
+  background-color: #f0f9ff;
+}
+
+.drop-zone.disabled {
+  opacity: 0.6;
+  cursor: not-allowed;
+}
+
+.hidden-input {
+  display: none;
+}
+
+.file-name {
+  font-weight: 600;
+  color: #303133;
+  word-break: break-all;
+}
+
+.hint {
+  font-size: 12px;
+  color: #909399;
+  margin-top: 8px;
+}
+
+.actions {
+  display: flex;
+  justify-content: center;
+  gap: 12px;
+  margin-top: 20px;
+}
+
+.btn {
+  padding: 10px 24px;
+  border: 1px solid #dcdfe6;
+  border-radius: 8px;
+  background-color: #fff;
+  color: #606266;
+  cursor: pointer;
+  font-size: 14px;
+  transition: background-color 0.2s, border-color 0.2s;
+}
+
+.btn:hover:not(:disabled) {
+  border-color: #409eff;
+  color: #409eff;
+}
+
+.btn:disabled {
+  cursor: not-allowed;
+  opacity: 0.6;
+}
+
+.btn.primary {
+  background-color: #409eff;
+  color: #fff;
+  border-color: #409eff;
+}
+
+.btn.primary:hover:not(:disabled) {
+  background-color: #66b1ff;
+  border-color: #66b1ff;
+}
+
+.message {
+  margin-top: 16px;
+  padding: 10px 14px;
+  border-radius: 8px;
+  font-size: 14px;
+  text-align: center;
+}
+
+.message.success {
+  background-color: #f0f9eb;
+  color: #67c23a;
+}
+
+.message.error {
+  background-color: #fef0f0;
+  color: #f56c6c;
+}
+
+.message.info {
+  background-color: #f4f4f5;
+  color: #909399;
+}
+
+.progress {
+  margin-top: 16px;
+  height: 6px;
+  background-color: #e4e7ed;
+  border-radius: 3px;
+  overflow: hidden;
+}
+
+.progress-bar {
+  width: 30%;
+  height: 100%;
+  background-color: #409eff;
+  border-radius: 3px;
+  animation: loading 1.2s infinite ease-in-out;
+}
+
+@keyframes loading {
+  0% {
+    transform: translateX(-100%);
+  }
+  100% {
+    transform: translateX(340%);
+  }
+}
+</style>

frontend/src/main.js

+import { createApp } from 'vue'
+import App from './App.vue'
+
+createApp(App).mount('#app')

frontend/vite.config.js

+import { defineConfig } from 'vite'
+import vue from '@vitejs/plugin-vue'
+
+export default defineConfig({
+  plugins: [vue()],
+  server: {
+    port: 5173,
+    proxy: {
+      '/api': {
+        target: 'http://127.0.0.1:5000',
+        changeOrigin: true,
+      },
+    },
+  },
+  build: {
+    outDir: 'dist',
+  },
+})