Cross Spectral Image Registration for License Plate Detection

Abstract

Geometrically aligning two or more images taken from various perspectives (multi-view registration), at various times
(multi-temporal registration), and via various senses (multi-modal registration) is known as image registration. While
images of the same scene may be obtained at different times or under different lighting conditions in multi-temporal
analysis, and by various types of sensors and imagers in multi-modal analysis, the images in multi-view analysis may
differ in translation, rotation, scaling, or more complex transformations, primarily because of camera positions. To
put it simply, image registration is the process of determining the spatial relationship between two or more sets of
images. Fourier analysis, cross correlation, Eigen value decomposition, moment matching, warping, procedural
method, anatomic atlas, internal landmarks, and many other algorithms are available for this. This study examines
methods based on correlation, features, and optimization.