Marvin-LiJunZan Image Enhancer Skill
Use this skill whenever you need to:
- Upscale low-resolution images: Enlarge blurry plots, charts, architecture diagrams, or experimental scans (2x, 4x).
- Denoise and Sharpen: Clarify blurry text, tick labels, or legend markers in academic figures.
- Batch restoration: Enhance figures before paper drafting, journal submission, or presentation slide compilation.
Architecture & Engines
The bundled Python script scripts/enhance_image.py supports multiple backends with automatic fallback:
- Engine 1 (Local Real-ESRGAN NCNN / PyTorch): Ultra-high quality AI super-resolution (Vulkan / CUDA / TensorRT).
- Engine 2 (Pillow / OpenCV High-order Lanczos & Unsharp Mask fallback): Guaranteed zero-dependency fallback that sharpens and upscales images instantly without GPU requirements.
Quick CLI Usage
Run with Python (supports Windows py -3.11 or standard python):
# Upscale 4x with automatic sharpening
py -3.11 skills/image-enhancer/scripts/enhance_image.py --input "path/to/blurry.png" --scale 4
# Specify explicit output path
py -3.11 skills/image-enhancer/scripts/enhance_image.py --input "path/to/blurry.png" --output "path/to/enhanced.png" --scale 4
Real-ESRGAN NCNN Integration
If Real-ESRGAN executable (realesrgan-ncnn-vulkan.exe) is placed in skills/image-enhancer/bin/ or in system PATH, the script automatically activates the neural network for publication-grade diagram & photo restoration.