Image Processing Exercises

This repository contains a series of foundational image processing exercises implemented in Python. The projects explore classical computer vision techniques such as image filtering, frequency-domain processing, segmentation, color analysis, and feature-based classification.

All exercises are implemented using standard scientific Python libraries (NumPy, scikit-image, Matplotlib), with a focus on understanding the underlying mathematical and algorithmic principles rather than relying on high-level black-box functions.

---

Repository Structure

• E1_image_merging/ – Manual image stitching and photometric adjustment
• E2_image_sharpening/ – Custom image sharpening using convolution filters
• E3_color_isolation/ – Color-based segmentation and grayscale transformation
• E4_superpixel_analysis/ – Superpixel extraction and clustering analysis
• E5_texture_classification/ – Feature-based texture classification

---

Project Overview

1. Manual Panorama Construction

A simple panorama is created by manually aligning two overlapping grayscale images.

Key steps:

• Load and convert images to grayscale
• Estimate translation alignment manually
• Merge images into a single panoramic output
• Apply gamma correction to match brightness characteristics

---

2. Image Sharpening via Custom Filtering

Implementation of a custom sharpening filter using Gaussian smoothing and identity kernels.

Key steps:

• Construct Gaussian blur filter
• Construct identity filter
• Combine filters to form a sharpening kernel
• Apply convolution to enhance image details
• Visualize both kernel and output image

---

3. Color-Based Segmentation

A script that isolates red regions in an image while converting all other regions to grayscale.

Key steps:

• Convert image to HSV color space
• Define a saturation-aware red mask (handling hue wrap-around)
• Apply selective grayscale transformation
• Reconstruct and visualize hybrid output

---

4. Superpixel Analysis and Clustering

Exploration of superpixels for image segmentation and structural simplification.

Key steps:

• Decompose image into superpixels
• Extract region-based statistics
• Represent image regions using feature vectors
• Apply clustering to group visually similar regions

---

5. Texture Classification

A basic texture classification pipeline using handcrafted features.

Key steps:

• Extract statistical features (mean, variance)
• Compute color histograms
• Analyze gradient direction distributions
• Classify textures (buffalo, elephant, zebra patterns)

Note: This exercise was implemented and tested in Google Colab.

---

Technologies Used

• Python 3.x
• NumPy
• scikit-image
• Matplotlib
• scikit-learn (for clustering and classification tasks)
• Google Colab (for Experiment 5)

---

Key Learning Outcomes

• Image filtering and convolution
• Manual image alignment techniques
• Color space transformations (RGB → HSV)
• Feature extraction for vision tasks
• Superpixel-based image representation
• Basic texture classification methods