Intel’s Automated Video Editing, Muvee, and MemoriesOnTV.

Those who know me understand that I started in the Media business as a video(tape) editor. Having started in the era of videotape, I have seen many things develop and one area that I’ve always been interested in is the concept of Automated Video Editing. When first hearing of this concept the vast majority of video editors pass the idea off as some sort of “tom-foolery” fearing either the creative aspects of their jobs will be taken away by some sort of robotic process or they are simply scared that their meal ticket will be taken away. From my experience nothing could be further from the truth.

I first came across the concept of Automated editing on a trip to Singapore when I ran into a couple of the countries leading researchers –Terence Swee and Dr. Pete Kellock– who were working in the area of audio and video mining. The results of their efforts is a great little program that I use on a PC called Muvee. Muvee is ideal for stringing together montages for use in just about any kind of TV program. It’s a great piece of software and it works on a number of levels. It operates by first scanning all of your video footage. It identifies faces, movement, shot brightness and a number of other parameters. It then scans the music that you want to use identifying its pace and breaks. Once the scan is completed you choose a style and it does all the editing for you. It works very well even with stills and has some great editing abilities. http://www.muvee.com/en/ Another well known program that I have tried is MemoriesOnTV which has been created for still images and is often associated with the Ken Burns effect. http://www.codejam.com/

Today I came across Intel’s latest research into video mining and a prototype Sports Highlight Viewer. It looks like another great editing tool. In its prototype form it looks like something that could start to pull the highlights automatically from a soccer game. A video was made by Intel to demonstrate how far this technology has come along with the advent of multicore processing.

It appears that the software can identify each player on the field, the ball and the net. I gather by logging the patterns of the movement and field positioning the software can identify and extract goals, penalty kicks and close plays. The extracted highlights then allow for a human operator to identify exactly what the play was and decide on its relevance for a sports highlight package. The demo goes on to show how much faster a highlight package can be generated by using multicore processing in the PC that creates the highlights. The video contains no audio so it’s pretty hard to distinguish exactly what the video demo is showing but there is a text description on the Intel research page that explains the video.

….brad….