Evaluating Feature-Based Image Registration Techniques for Aeroplane and Brain MRI Images using Projective Transformation

Image registration is required for the geometric alignment of two or more images acquired under varying conditions, viewpoints, times, or sensors by determining the optimal spatial transformation that aligns corresponding structures and features, thereby en- abling meaningful comparison, fusion, and analysis of the image data. Hence proposed an evaluation of eight feature-based image registration techniques for aeroplane images and brain MRI scans using projective transformation. The techniques examined in- clude SURF, FAST, BRISK, Harris, MinEigen, MSER, KAZE, and ORB. The research assesses each method’s effectiveness in detecting features and successfully matching them between image pairs. Results show significant variations in performance across techniques and image types. For aeroplane images, Harris demonstrated superior performance, detecting the highest number of features (2077 and 3017) and successfully matching 168. SURF and KAZE showed good results. In brain MRI registration the SURF performing best by matching 30 features out of 412 and 59 detected. Despite detecting numerous features, most techniques exhibited challenges in effectively matching features in MRI images. The proposed research can be utilized for medical diagno- sis accuracy and aviation safety through improved image registration techniques, benefiting healthcare outcomes and industrial inspection applications significantly.