An affine (or first-order) optic flow model has 6 parameters, describing image translation, dilation, rotation and shear. The class affine_flow provides methods to estimates these parameters for two frames of an image sequence.
The class implements a least-squares fit of the parameters to estimates of the spatial and temporal grey-level gradients. This is an extension of the well-known Lucas-Kanade method. The images are either sampled conventionally, on a rectilinear grid, or on a log-polar grid. In the latter case, the class may iteratively refine its estimates by moving the sampling grid to track the motion. Options to specify a region of interest and smoothing and sampling parameters are provided.
The file includes a demonstration of the class, and test images for this. The functions for smoothing images and estimating gradients may be useful independently, and log-polar sampling functions are included (and are available separately in submission 27023). |
