what should i change if i want to use user defined ROI (like the rect in LucasKanade) in the affine lucas-kanade

[wanghetongtt]

[ziliHarvey]

LucasKanadeAffine is used in dominant motion estimation to derive the transformation matrix M that warps a previous image to the current frame, which takes the entire image L_t as the template to be tracked in image L_(t+1), this approach is reasonable under the assumption that a majority of the pixels correspond to the stationary objects in the scene whose depth variation is small relative to their distance from camera.

In short, for LucasKanadeAffine, there's no rect since the entire Lt image is the template., you are welcomed to implement it as a LucasKanade version which takes into a L_t, L_(t+1), and a rect template.