Comics Upscaler

Comics Upscaler 是一个专门用于优化电子书（EPUB）图片质量的工具。它能够自动提取电子书中的图片，使用AI超分辨率技术提升图片质量，并重新打包成适合电子墨水屏阅读的电子书。

✨ 特性

• 🚀 支持EPUB格式的电子书处理
• 🖼️ 使用先进的AI超分辨率技术提升图片质量
• 📱 自动优化图片尺寸，适配电子墨水屏
• 📚 保留原始电子书的元数据和目录结构
• 🛠️ 支持批量处理多个文件
• ⚡ 多进程并行处理，提升处理速度
• 🧹 自动清理临时文件

🔧 系统要求

• Windows 10/11 64位
• Python 3.8-3.12
• 内存：建议8GB以上
• 存储空间：建议预留处理文件大小10倍以上的空间
• GPU要求：
  • NVIDIA显卡（推荐）：支持CUDA加速
  • AMD显卡：仅支持在Linux系统下通过ROCm使用
  • Intel显卡：暂不支持GPU加速
• Final2x-core（用于图片超分辨率处理）

📦 安装

方式一：直接下载（推荐）

1. 从 Releases 页面下载最新版本的压缩包
2. 解压到任意目录
3. 下载 Final2x-core 最新版本，解压并放置在程序目录下的 Final2X 文件夹中
4. 根据需要修改 config/settings.yaml 配置文件（参考下方配置说明）
5. 双击运行 ComicsUpscaler.exe

方式二：手动安装

1. 克隆仓库：

git clone https://github.com/[your-username]/comics-upscaler.git
cd comics-upscaler

2. 安装依赖：

pip install -r requirements.txt

3. 下载并配置Final2x-core：
   • 从Final2x-core Releases下载最新版本
   • 将项目解压放置在Final2X目录下

⚙️ 配置

在config/settings.yaml中配置以下参数：

# 临时目录
temp_dir: "temp"

# 目录设置
directories:
  # 输入目录路径
  input: "D:/电子书/"
  # 输出目录后缀（将添加到输入目录后）
  output_suffix: "_upscale"

# 超分辨率设置
upscale:
  # 使用的模型名称
  model_name: "RealCUGAN_Conservative_4x.pth"
  # 放大倍数
  scale: 4
  # 目标长边尺寸（适配电子墨水屏）
  target_long_edge: 3744
  # 并行处理的进程数
  num_processes: 4
  # 输出图片格式
  output_format: "JPEG"
  # 输出图片质量 (1-100)
  output_quality: 95

# EPUB设置
epub:
  # 是否调整到原始大小
  resize_to_original: false
  # 是否创建新的EPUB
  create_new: false
  # 是否创建适配电子墨水屏的EPUB
  create_eink: true 

超分辨率模型说明

• RealCUGAN-pro: 适合漫画、插画等内容，对线条和细节的保持较好
• RealESRGAN: 适合真实照片，对噪点的处理较好
• SwinIR: 新一代通用超分辨率模型，效果优秀但处理速度较慢
• EDSR: 经典模型，速度和效果均衡
• SCUNet: 适合处理复杂纹理的图像

🚀 使用方法

1. 将需要处理的电子书文件放入input目录
2. 运行程序：

python comics_upscale.py

3. 处理后的文件将保存在"input_upscaled"目录中

性能调优

• num_processes: 建议设置为CPU核心数的一半
• 如果内存充足，可以适当增加并行处理数
• GPU处理时建议将num_processes设置为1

📝 注意事项

• 处理过程中会占用较多磁盘空间，请确保有足够的存储空间
• 图片处理时间取决于图片数量、尺寸和电脑性能
• 建议先处理单个文件测试效果

❓ 常见问题

1. Q: 处理速度很慢怎么办？

   • A: 确保使用NVIDIA显卡并启用GPU加速
   • A: 可以尝试使用较轻量级的模型（如EDSR）
   • A: 减小目标分辨率或降低放大倍数

2. Q: 内存占用过高怎么办？

   • A: 减少并行处理数量
   • A: 降低处理图片的数量
   • A: 使用GPU处理时设置num_processes为1

3. Q: 支持哪些图片格式？

   • A: 支持jpg、png、webp等常见格式

4. Q: 为什么无法使用GPU加速？

   • A: Windows系统目前仅支持NVIDIA显卡的GPU加速
   • A: AMD显卡仅支持在Linux系统下通过ROCm使用
   • A: Intel显卡暂不支持GPU加速

致谢

特别感谢以下项目的支持：

• Final2x-core - 提供强大的图像超分辨率处理能力

---

Comics Upscaler

Comics Upscaler is a specialized tool for enhancing image quality in electronic books (EPUB). It automatically extracts images from e-books, uses AI super-resolution technology to improve image quality, and repackages them into e-books suitable for e-ink screen reading.

✨ Features

• 🚀 Support for EPUB format e-books processing
• 🖼️ Advanced AI super-resolution technology for image quality enhancement
• 📱 Automatic image size optimization for e-ink screens
• 📚 Preserves original e-book metadata and directory structure
• 🛠️ Support for batch processing multiple files
• ⚡ Multi-process parallel processing for improved speed
• 🧹 Automatic cleanup of temporary files

🔧 System Requirements

• Windows 10/11 64-bit
• Python 3.8-3.12
• Memory: 8GB or more recommended
• Storage Space: Recommend reserving 10x the size of files being processed
• GPU Requirements:
  • NVIDIA GPU (Recommended): Supports CUDA acceleration
  • AMD GPU: ROCm support on Linux systems only
  • Intel GPU: GPU acceleration not currently supported
• Final2x-core (for image super-resolution processing)

📦 Installation

Method 1: Direct Download (Recommended)

1. Download the latest version from the Releases page
2. Extract to any directory
3. Download the latest version of Final2x-core, extract and place it in the Final2X folder under the program directory
4. Modify the config/settings.yaml configuration file as needed (refer to configuration instructions below)
5. Double-click ComicsUpscaler.exe to run

Method 2: Manual Installation

1. Clone the repository:

git clone https://github.com/[your-username]/comics-upscaler.git
cd comics-upscaler

2. Install dependencies:

pip install -r requirements.txt

3. Download and configure Final2x-core:
   • Download the latest version from Final2x-core Releases
   • Extract and place the project in the Final2X directory

⚙️ Configuration

Configure the following parameters in config/settings.yaml:

# Temporary directory
temp_dir: "temp"

# Directory settings
directories:
  # Input directory path
  input: "D:/ebooks/"
  # Output directory suffix (will be added to input directory)
  output_suffix: "_upscale"

# Upscale settings
upscale:
  # Model name to use
  model_name: "RealCUGAN_Conservative_4x.pth"
  # Scale factor
  scale: 4
  # Target long edge size (for e-ink screen compatibility)
  target_long_edge: 3744
  # Number of parallel processes
  num_processes: 4
  # Output image format
  output_format: "JPEG"
  # Output image quality (1-100)
  output_quality: 95

# EPUB settings
epub:
  # Whether to resize to original size
  resize_to_original: false
  # Whether to create new EPUB
  create_new: false
  # Whether to create e-ink optimized EPUB
  create_eink: true 

Super-resolution Model Description

• RealCUGAN-pro: Suitable for comics and illustrations, better preservation of lines and details
• RealESRGAN: Suitable for real photos, better noise handling
• SwinIR: New generation universal super-resolution model, excellent results but slower processing
• EDSR: Classic model, balanced speed and effect
• SCUNet: Suitable for processing images with complex textures

🚀 Usage

1. Place the e-books to be processed in the input directory
2. Run the program:

python comics_upscale.py

3. Processed files will be saved in the "input_upscaled" directory

Performance Tuning

• num_processes: Recommended to set to half of CPU cores
• Increase parallel processing if memory is sufficient
• Set num_processes to 1 when using GPU processing

📝 Notes

• Processing will occupy significant disk space, ensure sufficient storage
• Image processing time depends on image quantity, size, and computer performance
• Recommended to test with a single file first

❓ FAQ

1. Q: What to do if processing is too slow?

   • A: Ensure NVIDIA GPU is used with GPU acceleration enabled
   • A: Try using lighter models (like EDSR)
   • A: Reduce target resolution or scale factor

2. Q: What to do about high memory usage?

   • A: Reduce number of parallel processes
   • A: Reduce number of images being processed
   • A: Set num_processes to 1 when using GPU

3. Q: What image formats are supported?

   • A: Supports jpg, png, webp, and other common formats

4. Q: Why can't I use GPU acceleration?

   • A: Windows systems currently only support NVIDIA GPU acceleration
   • A: AMD GPUs only supported through ROCm on Linux systems
   • A: Intel GPUs currently do not support GPU acceleration

Acknowledgments

Special thanks to:

• Final2x-core - Providing powerful image super-resolution processing capabilities