Fill out the black area in video stabilization

edit

Hi everyone,

I am researching about video stabilization. In final step, I need to fill out the black area in video frame to acquire pleasant looking. I did it by normal way - use mosaic from previous and next frame of current frame to fill out (Just get the pixel value in valid area of previous/next frame into black area (undefined area)) But it is seem not good and deformed.

Here is the result: link text

I found some ways to improve it. The paper also use mosaicking, but so confused to understand it link text

Could you explain or suggest better way to fill out the black area, please?

Thank you in advance.