A pattern search method for image registration
This paper presents a pattern search method for transformation function search in automatic image registration. In this search method, affine transformation parameters are represented as a 5-parameter vector and the search proceeds toward the direction resulting in higher similarity values between the reference and sensed images. Experiments show that this method can successfully find an optimal affine transformation function for edge images generated by computing the local standard deviation of the images. In addition, a series of overlapped aerial images of the Mississippi Delta area are successfully registered into a large reference image automatically by employing this search method.
Lecture Notes in Computer Science
Seyfarth, B. R.
(2005). A pattern search method for image registration. Lecture Notes in Computer Science, 3514, 664-670.
Available at: http://aquila.usm.edu/fac_pubs/2939