Chromatic Resurrection is an image restoration and colorization project designed to transform grayscale flower images into visually realistic colored outputs.
The system applies machine learning-based pixel prediction techniques to reconstruct missing color information while preserving image structure and details.
Convert:
⚫ Black & White Images
into
🌸 Realistic Colored Images
using AI-assisted image processing methods.
✔ Grayscale → Color image transformation
✔ Image restoration
✔ Enhancement of visual appearance
✔ Side-by-side comparison view
✔ Web interface for interaction
✔ Prediction-based color reconstruction
Current implementation uses:
Patch-based Multiple Linear Regression
for grayscale-to-color prediction.
Future enhancements proposed:
- GAN
- CNN Encoder–Decoder
- U-Net
- Pix2Pix
- Attention-based models
Core:
Python • OpenCV • NumPy • Scikit-learn
Visualization:
Matplotlib
Web:
Flask • HTML • CSS • JavaScript
Dataset source:
Kaggle flower image colorization datasets
Approx training size:
Large-scale image dataset (18~GB-level training data)
Possible datasets:
- Flower image colorization
- Image restoration datasets
Project evaluated using:
| Metric | Purpose |
|---|---|
| MSE | Error measurement |
| R² Score | Regression quality |
| PSNR | Image reconstruction quality |
Potential future expansion:
✔ GAN-based colorization
✔ CNN image reconstruction
✔ U-Net segmentation pipelines
✔ Improved realism generation
Current status:
Local Flask deployment
Public deployment:
Not yet available
This work explores:
Computer Vision
Image Restoration
Machine Learning
AI-assisted media reconstruction
Mugash Priyan U
Ananthanarayana M
Rohith S
Completed as an academic AI/ML project with scope for deep learning enhancement.
⭐ If you found this project interesting, consider starring the repository.