Computer Vision & Image Processing Projects

This repository contains a collection of computer vision projects and laboratory exercises implemented in Python using OpenCV and scikit-image. The notebooks cover fundamental to advanced image processing techniques ranging from basic binarization to feature extraction and image stitching.

📂 Repository Structure

1. Image Binarization

• File: 01_Binarisation.ipynb
• Description: - Investigation of grayscale and color image binarization methods.
  • Focuses on separating objects from the background and isolating individual objects.
  • Includes experimental analysis of parameter selection for thresholding algorithms (beyond simple cv2.threshold usage).

2. Edge Detection & Histogram Analysis

• File: 02_Edge_Detection.ipynb
• Description:
  • Parts 1-3: Implementation of edge detection algorithms.
  • Part 4: Global operations focusing on histogram transformations and further binarization techniques.

3. SLIC Superpixel Segmentation

• File: 03_SLIC_Segmentation.ipynb
• Description:
  • Segmentation of images (e.g., 'tabby', 'caries') using the SLIC (Simple Linear Iterative Clustering) algorithm.
  • The goal is to obtain uniform superpixels that contain either only parts of the object or only the background.
  • Analysis of the algorithm's behavior with varying numbers of superpixels and sigma parameters.
  • Utilization of skimage.segmentation.slic and boundary visualization.

4. Classic Object Detection (Morphology & Blobs)

• File: 04_Classic_Detection.ipynb
• Description:
  • A pipeline for generating synthetic training data for amodal segmentation.
  • Tasks: Detection and segmentation of specific objects (larvae) from source images.
  • Pipeline: Color binarization -> Connected Components (cv2.findContours) -> Blob Analysis.
  • Filtering: Removal of noise, merged instances, and non-target objects.
  • Measurements: Estimation of object properties such as length (via skeletonization), thickness, and shape classification (I-shape, S-shape, C-shape).

5. Hough Transform (Lines & Circles)

• File: 04_Hough_Detection.ipynb
• Description:
  • Application for analyzing snooker positions.
  • Goal: Detect balls and table cushions (bands).
  • Methods: - Edge detection preprocessing.
    • Hough Line Transform (cv2.HoughLines) for detecting cushions.
    • Hough Circle Transform (cv2.HoughCircles) for detecting balls.
  • Includes parameter calibration and visualization of detected geometry on the input images.

6. SIFT Feature Detection (Research)

• File: 05_SIFT_POI_Detection.ipynb
• Description:
  • Research and basic usage of the SIFT (Scale-Invariant Feature Transform) algorithm via cv2.SIFT_create.
  • Detection of KeyPoints (detect) and computation of Descriptors (compute).
  • Visualization of features using cv2.drawKeypoints.

7. Image Stitching (SIFT Implementation)

• File: 05_SIFT_HARD_EXAMPLE.ipynb
• Description:
  • A practical project focused on Image Stitching (Panorama creation).
  • Uses a custom dataset of a static scene captured from different offsets and angles.
  • Workflow: 1. Detect SIFT keypoints and descriptors. 2. Match features between the base image and offset images. 3. Calculate geometric transformations (Homography). 4. Merge images to create a unified view.