📄 Document Deweathering Engine

An unsupervised machine learning application designed to restore weathered, stained, or degraded documents. This project implements Robust Principal Component Analysis (RPCA) using the Inexact Augmented Lagrange Multiplier (IALM) algorithm with Adaptive Regularization.

🚀 The Core Engineering

Most restoration tools use static filters. This engine uses Matrix Optimization to solve document degradation. It treats an image as a matrix $D$ and decomposes it into:

• Low-Rank Matrix (A): The background interference, illumination, and stains.
• Sparse Matrix (E): The sharp text and handwriting.

The system solves the following optimization problem: $$\min_{A,E} |A|_* + \lambda |E|_1 \quad \text{s.t.} \quad A + E = D$$

🧠 New: Adaptive Regularization

Unlike standard RPCA implementations, this engine uses a Dynamic Heuristic to adjust the regularization parameter $\lambda$ based on the input's noise density.

• Noise Analysis: The backend calculates the standard deviation ($\sigma$) of the input matrix.
• Scaling: It dynamically scales the penalty parameter $k$ between 0.4 and 0.8.
• Impact: This ensures that clean documents aren't "over-processed" while heavily stained documents receive aggressive restoration.

🛠️ Technical Stack

• Backend: Python, FastAPI, NumPy, OpenCV (Adaptive SVD-based optimization)
• Frontend: React, Vite, Tailwind CSS (Glassmorphism UI)
• Version Control: Git & GitHub

📂 Project Structure

• /backend: The Python math engine and FastAPI server.
• /frontend: The React web application interface.
• /assets: Project screenshots and demo media.

⚙️ How to Run

1. Start the Backend:

   cd backend
   uvicorn main:app --reload