Panoramic stitching

• Main Author: Lek, Samuel Yong Chuan
• Other Authors: Chua Chin Seng
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/75669
• Institution: Nanyang Technological University

Panoramic stitching is an image processing technique that involves combining two or more images with overlapping section to create an image with a larger field of view. There are a myriad of applications for panoramic stitching ranging from virtual reality to interior design. This project however, does not focus on the application of panoramic stitching but instead on realizing the fundamental theories of panoramic stitching and demonstrating these theories in codes and visual displays. In this project, we will be working on CodeBlocks with the OpenCV library. We will be looking at 2 feature point extraction techniques, namely the Harris Corner Detector and the Speeded-Up Robust Feature (SURF) Detector. Both detectors aim to capture feature points robustly such that the feature points under various lighting conditions and rotation and translation will be consistent. We will be using these detectors for feature matching and image stitching and in doing so, we will compare the reliability of both detectors. For feature matching, we will be demonstrating techniques of normalizing a pixel to compensate for brightness, error computation based on Sum of Squared Difference (SSD) of feature point patch and error computation based on total mean and median error of matched points. In doing so we will also determine and explain the effectiveness between mean and median error computation. For image stitching, we will be displaying the correlation between two images for easy verification of matching feature points and we will be performing image combination based on the best match feature point pair. Finally, will be looking at different test results by running our program against different images (ideal web, ideal camera, practical camera) and we will be providing an analysis of these results and explain the inadequacies in our program.