This repository showcases my implementations of various 3D Computer Vision applications and techniques completed while undertaking the ECE661 coursework. It encompasses a range of topics, including:
Image Projection and Rectification: Developed expertise in image manipulation by implementing homographies for image projection onto other images. Applied advanced techniques, such as point-to-point correspondence methods and dual degenerate conic, to rectify distorted images and ensure accurate alignment.
Image Mosaicing and Panorama Generation: Demonstrated proficiency in computer vision algorithms by using the RANSAC algorithm for outlier detection and the Levenberg-Marquardt algorithm for non-linear least squares minimization. Successfully created stunning panoramic images by seamlessly combining multiple images, resulting in visually coherent and professional-grade panoramas.
Interest Point Detection and Feature Extraction: Acquired strong knowledge of interest point detection techniques by implementing Harris corner detection, as well as SIFT/SURF algorithms using SSD and NCC similarity metrics. Extracted robust and distinctive features from images, enabling accurate object recognition and matching.
Image Segmentation and Contour Extraction: Developed expertise in image segmentation by implementing the Otsu algorithm based on RGB channel values. Utilized advanced techniques such as dilation and erosion for texture-based segmentation to extract contours accurately. Achieved precise object boundaries for further analysis and processing.
Image Classification with Texture Characterization: Showcased proficiency in image classification techniques by implementing texture characterization using LBP Histogram and Gram matrix feature vectors. Achieved an outstanding accuracy rate of 98% in image classification by employing Gram matrix-based classification, demonstrating strong analytical and machine learning skills.
Camera Calibration and 3D Reconstruction: Gained practical experience in camera calibration using Zhang's Algorithm, utilizing Canny edge detection for corner detection and applying the LM algorithm for corner projection with radial distortion correction. Employed projective stereo rectification techniques for 3D reconstruction of uncalibrated camera images, demonstrating a solid understanding of geometric transformations and 3D reconstruction principles.
Facial Recognition with Dimensionality Reduction: Demonstrated proficiency in facial recognition algorithms by implementing PCA and LDA dimensional reduction techniques. Employed the K-Nearest Neighbors (KNN) algorithm for accurate identification and achieved impressive results in facial recognition, highlighting strong pattern recognition and machine learning skills.