“The just noticeable difference of transition durations” by Wang and Bodenheimer

This work reports results on the noticeability of transitions between two segments of motion. Linear interpolation is often used as a method of generating such transitions because it is efficient and quick. [Wang and Bodenheimer 2004] reported methods to automatically compute such durations. In this work, we refine and extend their study of how noticeable the blend length is to viewers. There is no generally accepted standard for generating or specifying a transition in the graphics community. Instead, two techniques predominate. Some researchers, e.g., [Rose et al. 1998] specify transitions using a start frame in motion 1, an end frame in motion 2, and a blend length d. We call this method a “start-end” specification. Other researchers, e.g., [Kovar and Gleicher 2003] specify a transition from frame i to frame j meaning that frames i and j are the midpoint in a transition of length d frames, i.e., the 50% blend points. We call this method a “center-aligned” specification. There are advantages and disadvantages to each of these two methods.

1. Kovar, L., and Gleicher, M. 2003. Flexible automatic motion blending with registration curves. In 2003 Symposium on Computer Animation, 214–224.
2. Rose, C., Cohen, M., and Bodenheimer, B. 1998. Verbs and adverbs: Multidimensional motion interpolation. IEEE Computer Graphics and Applications 18, 5, 32–40.
3. Wang, J., and Bodenheimer, B. 2004. Computing the duration of motion transitions: an empirical approach. In 2004 Symposium on Computer Animation, 335–344.
4. Wetherill, C. B., and Levitt, H. 1965. Sequential estimation of points on a psychometric function. Brit. Jour. of Math, and Stat. Psych. 18, 1–10.