A comprehensive Computer Vision pipeline implemented in Python. This project focuses on building core image processing algorithms from scratch (using NumPy) to demonstrate a bottom-up understanding of pixel manipulation, feature extraction, and geometric transformations. For a more detailed analysis check the PDF report!
- Part A: Convolution & Filtering: Manual 2D convolution implementation. Includes custom Gaussian and Mean filters, and noise simulation (Gaussian, Salt & Pepper, Poisson).
- Part B: Edge Detection: Custom Canny Edge Detector implementation (gradient magnitude, non-maximum suppression, and hysteresis thresholding).
- Part C: Corner Detection: Implementation and comparison of Moravec and Harris corner detection algorithms, including multi-scale analysis using Gaussian pyramids.
- Part D: Document Rectification: Automated perspective correction (document scanning) using Harris corners, geometric quadrant heuristics, and Homography.
ImageRectificationToolkit_CV_2025_Ioannis_Petrakis.ipynb- The main Jupyter Notebook containing all code and experiments.assets/- Directory for input test images (e.g., book.jpg, building.jpg, etc.).results/- Directory containing saved output visualizations.
Clone the repository:
git clone https://github.com/DONGIOVANNI7/ImageRectificationToolkit_CV.git
cd ImageRectificationToolkit_CV