Computer Vision - Transformations

This repository contains a complete set of image processing and computer vision exercises developed for Computer Vision university course at U-tad (Centro Universitario de Tecnología y Arte Digital). All tasks were implemented in a single Jupyter notebook, supported by an img/ directory containing test images.

The work covers a wide range of transformations and filtering techniques, including contrast enhancement, denoising, Gaussian filtering, edge detection, feature extraction and connected-components analysis. Both manual implementations and OpenCV-based methods are explored throughout the notebook.

Contents Overview

Problem 1: Local Contrast Enhancement

• CLAHE using OpenCV
• Manual CLAHE implementation

Problem 2: Global Contrast Adjustment

• Single and multiple clipLimit experiments
• Manual multi-clipLimit contrast enhancement

Problem 3: Noise Reduction and Filter Evaluation

• Application of linear and nonlinear denoising filters
• MSE and PSNR quantitative evaluation
• Median-filter section analysis

Problem 4: Advanced Denoising via NLMeans

• NLMeans applied to grayscale images
• Quantitative comparison and section analysis

Problem 5: Gaussian Smoothing and Separable Convolution

• Construction of the Gaussian kernel
• Horizontal and vertical separable convolution
• Manual Gaussian filtering vs OpenCV implementation

Problem 6: Edge-Preserving Smoothing

• Theoretical basis of the Kuwahara filter
• Qualitative interpretation

Problem 7: Edge Detection and Second-Order Operators

• Laplacian of Gaussian (LoG)
• Canny edge detection

Problem 8: Object Detection Through Connected Components

• Size reduction + Canny preprocessing
• Connected-components extraction
• Comparison with alternative approaches

Problems 9-10: Scale-Space Keypoint Detection and Matching

• SIFT keypoint detection
• Orientation assignment
• Descriptor construction
• Feature matching with BFMatcher

Repository structure

/img        # Test images
computer_vision_transformations.ipynb
README.md

Note

All results, plots, explanations, and comparisons are contained inside the notebook.

Techniques demonstrated

• Contrast enhancement (CLAHE)
• Noise reduction and smoothing filters
• Gaussian filtering (manual & OpenCV)
• Laplacian of Gaussian
• Canny edge detection
• Connected components
• SIFT: detection, description, and matching
• Quantitative measures: MSE, PSNR, entropy