You are here:

Video editing by modelling

Image editing packages have become powerful and flexible tools for manipulating visual information. Efforts to provide the same functionality and ease of use within video editing software, however, have so far failed. This is in part due to the fact that many basic editing operations, such as moving an object from one video stream to another, require information about the shape of the objects in the scene. Operations which are common in image editing programs such as synthetic object insertion, object removal, re-lighting and shadow insertion have the same limitation. There are also a number of video-specific operations, such as re-shooting the video with a new camera path, or with changed camera settings, which require an understanding of the shape of the scene viewed.

The problem

The difficulty in performing even seemingly simple video editing operations such as cutting and pasting between sequences is that just shifting the pixels doesn't work.

Effective video editing requires 3D models. To cut and paste between sequences you need a model of the object to be transferred, and a model of the destination scene. Removing an object from a video sequence similarly requires a 3D model in order to determine what the deleted object might be replaced with.

More complex compositing operations such as depth of field effects, smoke, or fog, also require 3D models of the scene viewed, models which VideoTrace generates quickly and easily.

Cutting and pasting between sequences

Using VideoTrace, these types of video processing operations become simple.

Cutting and pasting objects between sequences requires that we model both the object to be moved, and the scene in the destination video. We perform this modelling with VideoTrace. More information on VideoTrace is available here.