Image Registration using Shi-Tomasi and SIFT
Sanika S Patankar, Sara Ganesh Kadam, Aishwarya Jadhav, Maitrayee Gore · 2023
The procedure of merging data from various sources into a single reference system/ reference frame is known as ‘Image Registration’. The data can be images taken from the sensors, multiple different images taken from different angles, images from different lapses of time, images from a different point of view/frame of reference, etc. This type of data is used in numerous applications such as medical imaging, computer vision, target detection in the military, analyzing and processing of satellite data, etc. To merge and differentiate this data obtained from various different sources, image registration is used. This paper introduces two algorithms well-balanced together for image registration. The two algorithms used are Shi-Tomasi Corner Detection Algorithm and Scale-Invariant Feature Transform (SIFT) Algorithm. The Shi-Tomasi Corner Detection Algorithm issued by J.Shi and C.Tomasi in the year 1994 is used for the corner detection in the reference image. The Scale-Invariant Feature Transform (SIFT) Algorithm is used for searching, outlining and equivalenting the corners/features in the images. The average error obtained after the estimation of the TRS for, Translation of (500,500), Rotation of angle -10, and Scale of 0.7 is $\theta=0.18\%,\Delta x=17.14\%,\Delta y=7.69\%,S=0.01\%$, and the average error obtained after the estimation of the TRS for, Translation of $(500,500)$, Rotation of angle 10 and Scale of 0.8 is $\Theta=0.14\%,\Delta\mathrm{x}=10.76\%,\Delta\mathrm{y}=7.95\%,\mathrm{~S}=0.01\%$