This paper proposes one method of doing image morphing. The authors propose using lines to do the warping. The idea is pretty simple. Drawing one line and its correspondence defines a coordinate system on both source and target images. For any pixel, we can calculate its coordinate relative to the line, and warping it accordingly. The authors note that it is better to have the perpendicular length be irrelevant of the line segment length, which means we can't do true scaling of images using this technique.  
  
Having more than one line only complicates things slightly. Now, each line has an influence. The influence dies out as we move away from the line. The location of a pixel is determined by taking the weighted average of its positions relative to all lines. Crossing lines seem to be a problem! So, the authors just decided to just make it a taboo! This doesn't seem much of a constraint anyway. With many lines, there is still no way to define uniform scaling! If we have 2 perpendicular lines,Scaling both lines causes distortions in the resulting image.  
  
Clearly, one of the disadvantages here is global control. Deciding to draw lines for eyebrows can end up disrupting how the mouth morphs and so on. Hopefully, though, this is not noticed because the disruption is very small.   
  
Another disadvantage is the running time. Adding one more line on the eye brows, increases the processing time for pixels near the chin! This is not a very desirable thing.  
  
Anyway, with all that being said, I believe we can easily implement their method as-is on today's computers and have it running at interactive rates for pretty large images.  
  
Overall, this was a nice read. It explains things very intuitevely, and I liked reading it.