Towards Collaborative Video Authoring

1
Towards Collaborative Video Authoring
University of St.Petersburg
Boris Novikov borisnov_at_acm.org
Oleg Proskurnin olegpro_at_acm.org
2
Video Authoring

• Post-production became digital

• Non-linear editing systems have presented new
  concepts
• video clip as an independent unit of raw material
• timeline to represent the time flow
• altering and assembling clips as a major part of
  the editing process
• rendering as a way of building the final video
  production

• Still there is no support for collaborative video
  authoring

3
Transactional Approach to Cooperation

• Ensures consistent exchange and sharing of
  information by taking into account application
  semantics

• Cooperative activity framework CoAct offers
• private workspace for each user involved
• common workspace to represent the state of team
  work
• activity histories for modelling the working
  process
• history merging algorithm to achieve cooperation

• Merging is guided by compatibility relations of
  editing operations

4
Towards a New Data Model

• Basic requirements
• compliance with video authoring concepts
  including explicit support for the most common
  editing operations
• good commutativity properties of these operations
• additional history maintenance mechanism to
  facilitate utilization of CoAct cooperation
  techniques

• Our approach
• assume video segments as basic units of the media
  stream
• consider operations on different segments as
  commuting
• provide a tree-based structure for arranging
  video segments and maintaining activity histories

5
Concurrent Video Video Segments

• Each video segment consists of
• a frame sequence referencing a contiguous block
  of video
• a set attribute-value pairs describing proper
  interpretation of raw data at the presentation or
  rendering level

• Attributes support implementation of
  non-destructive editing operations

6
Concurrent Video Video Activity Tree

• Unique identifiers associated with video segments
  provide background for efficient editing
  operations with state-independent commutativity
  relations

7
Concurrent Video Insertion

• Insertions carried out within different segments
  commute

8
Concurrent Video Deletion
raw videos
timelines

• Any two removals commute
• Deletion acts as an inverse for some insertions
• Additionally local activity history clean-up is
  performed

9
Concurrent Video Moving

• Moving is not equivalent to the sequence of
  appropriate deletion and insertion operations
  from a semantic point of view

• Special moving operation offers
• behavior similar to deletion-insertion pair
• constancy of the unique identifier of the segment
  being moved
• commutativity with the editing operation

10
Concurrent Video Editing

• General operation for high-level modeling of
  various modifications of image data within
  particular video clips

• Editing operation offers
• altering of video clip data within any valid node
• modeling of special effects and transitions by
  means of video segments attributes

11
Concurrent Video Commutativity Relations

• Forward commutativity is used for detecting
  conflicts between operations coming from
  different histories during merging

• Backward commutativity is different
• it determines dependencies between operations
  within a single history
• it allows us to identify closed subhistories,
  which have no external dependencies and thus
  represent consistent units of work
• consistent units of work can be exchanged between
  cooperating users separately from each other

12
Concurrent Video Consistent Units of Work

• Extraction of a consistent unit of work according
  to the users selection is performed by means of
  the partial hierarchical order which naturally
  captures dependencies between operations
• Totally ordered closed subhistory for the above
  example
• (EditA, EditX, Move, EditB, Insert, EditK, EditL)

13
Concurrent Video Merging

• Forward commutativity properties of editing
  operations enable high concurrency among
  co-workers actions

Operations Edit(ID, VS) Insert(VS, ID, POS) Delete(ID) Move(MID, ID, POS)
Edit(id, vs) id ? ID
Insert(vs, id, pos) id ? ID id ? ID
Delete(id) id ? ID id ? ID true
Move(mid, id, pos) id ? ID mid ? ID ?id ? ID mid ? ID ?id ? ID mid, id ?MID, ID ?
14
Conclusions and Further Work

• Concurrent video ...
• is a formal basis for collaborative video
  authoring environments
• satisfies basic requirements of authoring
  applications as well as the needs of multi-user
  systems
• is applicable to stream data in general, e.g. to
  audio data

• Probable directions of futher research
• development of intra-segment editing operations
• investigation of versioning support