A new multiwavelets-based motion estimation

This paper presents a novel motion estimation method based on the multiwavelets analysis. An omnidirectionally balancing multiwavelets for vector transform is used to estimate accurate dense optical flows in an image sequence. This multiwavelets motion model is particularly useful in estimating optical flows from the image sequence. In the gradient-based method, the computation of the motion field depends on minimizing the energy function and smoothness constraints. The proposed approach takes advantage of the nature of the multiwavelets theory, which can simultaneously possess desirable properties to model the vector fields and image related functions. Each flow vector and image function is represented by a linear combination of the multwavelets basis functions. The estimates is used a coarse-to-fine hierarchical strategy to produce the accurate motion field that are robust to noise and perspective deformation.