Automatic Video Authoring with Media Analysis by pengxiang


									Automatic Video Authoring with
       Media Analysis


       Chen-hsiu Huang
     Advisor: Dr. Ja-Ling Wu
 Problem  formulation
 Solutions and framework
 Expectations
 Status report
 Demonstration
 Questions and discuss
             Problem Formulation
   The digital video capture devices such as DVs are made
    more affordable for end users.
   It’s interesting to shoot videos but frustrating for editing
   There’s still a tremendous barrier between amateurs (home
    users) and the powerful video editing software.
   Finally people leave their precious shots in piles of DV
    tapes without editing and management.
     Type I home video: memorial
     Type II home video: recreational
   According to a survey on
    DVworld*, the relations       Video length   Review times
    between the video length      >= 1 hr        1 or 0
    and how many times will
    user review them after        30 min ~ 1 hr 2 ~ 3
    days:                         15 ~ 30 min    5 ~ 10
   People are inpatient for      5 ~ 15 min     >= 10
    videos without scenario or
                                  <= 5 min       You take it out and
    voice-over, especially for                   watch it when you
    those with no music.                         think about!
   Video clips with no more      *
    then 5 minutes are best for
    human’s concentration.
        Solutions and Framework
   A consumer product called “muvee autoProducer” has
    been announced to ease the burden of professional video
   The application scenario is quite simple:
     Pickup a video  Choose your music  Produce musical video!
   Although there are commercial products in the market, only
    few academic publications related.

   Goal: To achieve the near or beyond quality in the similar
    application scenario with the content-analysis technologies
    developed in multimedia domain.
 Brightness        Framework of AVAMA system

                                                            Currently finished
 Input music         cutting


                   Shot change                                      Output Video
                  Scene change
Input video

 Human face                                            Audio rhythm &
  Flash light               Scene selection           Video motion/color
Motion strength            Key shot selection          synchronization
Color variance
   Quality results
   Time & space complexity consideration
     Must not take too longer to make a video
     Must not consume too much memory
   Complete consumer software implementation
     Deal with MPEG-1/2 videos and MPEG-1 Layer 1/2/3 audios
      directly without pre-processing
     Simple application scenario to produce videos
   The need of different profile?
   Is semi-automatic necessary?
                      Status Report
   Audio Cutting:
     Cutting with zero-crossing rate (not so good)
     Cutting with dramatic volume change (great results)
     Brightness, bandwidth, ...
   Video Segmentation:
     Shot change detection by pixel MAD
     Detect scenes with flash light event
     Detect fast motion scenes
         Longer fast motion scenes means author’s tracing some objects
         Short fast motion scenes may due to
     Detect human face information (time complexity?)
           Questions and Discuss
   Any comments are welcomed.
   Special thanks for Mr. 劉嘉倫, for his videos and
    suggestions. 
   Thanks friends in DVworld who provide lots of ideas and
Create Music Videos using Automatic Media
     Analysis, ACM Multimedia, 2002
Let’s see its demo!


To top