Image registration based on KLT feature tracker in image mosaicing application
Shu Zhilong, Qiuqi Ruan · 2002
Image mosaicing has many applications, such as generating panoramas, video surveillance, MPEG-4 encoding, image based rendering, etc. The critical technique of image mosaicing is image registration. In this paper a new, efficient and robust method based on phase correlation and KLT feature tracker is developed for image registration and applied in image mosaicing. The KLT feature tracker refines the result of phase correlation and gives a better initial value for the Levenberg-Marquard (L-M) algorithm, so a global optimal solution can be obtained.