Application of Stochastic Adaptation in Block Method for Estimating Image Sequence Deformation Field

Abstract

The paper proposes a block method for estimating the deformation field for image sequences using a stochastic adaptation procedure. As a target function, the paper considers mean square inter-frame difference and interframe correlation coefficient. The developed algorithm has a high noise resistance and allows one to reduce the influence of global inter-frame geometric changes. Also, by adjusting the size of the blocks, it makes it possible to remove or highlight moving objects of small size (rain, snow, falling leaves, etc.).