Segment Analysis for Nonparametric Image Registration of Spaceborne Imagery of Farmland
Ruchi R. Mittal, Antony Mikalaevich Kazlouski, Ajay Rana · 2022
Segment analysis is a daunting problem in image processing. There is interest in its use along with nonparametric image registration to align meaningful data of multitemporal remote sensing imagery, which is critical for the successful longterm monitoring of a specific land feature. This study deals with the application of the aforementioned technologies to effectively align segments of different spaceborne images. On the one hand, the mathematical apparatus for segment analysis is leveraged to extract significant information. The concept of a contour is used. The improved model of a segment is introduced, as well as the formal description of a composite object is enhanced. On the other hand, nonparametric registration techniques for optimizing an objective function, quantifying the similarity measure, and regularization term provides a flexible approach for image registration. As a result, the registration approach is based on segment analysis, multilevel registration, curvature regularisation, and normalised gradient field distance measure. A formal description of the segment has been developed. The technique can register segments of remote sensing imagery precisely. This is proved by experiments.