You are on page 1of 47

The Visual Systems Lab

The Evolution of Film Editing Technique


and its Implications to the Parsing and
Summarization of Motion Pictures

John Mateer
Dept of Electronics
Dept of Theatre, Film & TV
(011 44) 1904 433 245
jwm10@ohm.york.ac.uk
The Visual Systems Lab

Overview
™ Overview of Shooting Related to Editing
™ Evolution of Grammar and Techniques
™ Relevance to Automated Parsing
™ Benefits to Summarisation
™ Issues and Obstacles
™ Conclusions
The Visual Systems Lab

Shooting in Relation to Editing


™ Director controls what is shot
š ‘Surrogate audience member’
™ Composition attributes affecting editing
š Subject Position
š Framing
š Lead space
š Eye line
The Visual Systems Lab

Subject Position within Frame

1/3rd

2/3rds

Camera Left Camera Right

Saving Private Ryan ©1998 Paramount


The Visual Systems Lab

Framings
The Visual Systems Lab

Lead Space

Saving Private Ryan ©1998 Paramount


The Visual Systems Lab

Eye Line

Saving Private Ryan ©1998 Paramount


The Visual Systems Lab

Shooting and Editing Scenes


™ Scene Structure
š A clear beginning, middle and end
š Not a resolution of story but of moment
™ Coverage
š All shots taken to record the scene
š Each new shot should advance story
• What does the audience need to see?
™ Processed raw footage called “Rushes”
The Visual Systems Lab

Choosing Shots
™ Variation in composition and angles will
determine how shots “cut”
™ Continuity
š In Setting and Space
š In Action
š In Camera motion
š The “Line”
The Visual Systems Lab

Thomas Crown Affair (1968)

Clip from the bank manager scene


The Visual Systems Lab

Relationship of Time to Editing


™ Time
š Running time – physical duration of program
š Audience time – perceived duration
š Story time – local duration within program
š Action time – event duration
™ Flow of time controlled through “beats”
š In action, camera movement or transitions
š Enable breathing space, emphasis
The Visual Systems Lab

Evolution of Film Grammar


™ Pre-1900: Birth of Cinema
™ 1900-1910: Evolution of Continuity
™ 1910-1920: Beginning of Common Language
™ 1920-1930: Conceptual Languages Emerge
™ 1930-onward: Expansion of established
grammars, development of new ones
The Visual Systems Lab

Pre-1900: Birth of Cinema


™ Camera not mobile
™ Requirement of light very high
™ Theatrical Presentation
š Fourth Wall model
š Audience kept distant
™ No transitions
™ Simply to record a constrained event
The Visual Systems Lab

1900-1910: Notion of Time


™ Long takes
™ Straight cut transitions emerge
™ No line but evolution of continuity
š entrance and exits from correct side
š action consistent between adjacent shots
™ Initially real time (no ellipsis)
™ Location work just starting
The Visual Systems Lab

Yorkshire Egg Collectors (1908)

Clip of ‘climmers’ gathering edge


from cliff faces in Flamborough
From the Yorkshire Film Archives
The Visual Systems Lab

1900-1910: Continuity Evolves


™ Linear time but in multiple locales
š very beginning of ellipsis
™ Variation in framing dictated by location
™ Audiences began to understand nature of
film and basic continuity
™ Novelty drives filmmaking
The Visual Systems Lab

The Thieving Hand (1908)

Clip of the hand being pawned,


stealing from the pawn shop
and the police being summoned
The Visual Systems Lab

1900-1910: Continuity Evolves


™ Early notion of ‘Line’ through Eye Line
š Lead space used to guide viewer’s attention
š Continuity of space still inconsistent
š Still ‘fourth wall’ model
™ Technical innovations enhance technique
š Stop motion photography, Multiple exposure
š Variation in recording frame rates
™ Audience sophistication drives innovation
The Visual Systems Lab

Princess Nicotine (1908)

Clip of the stop-motion animation of


cigar being created, man in living room
smoking and tormenting the Princess
The Visual Systems Lab

1910-1920: Common Language


™ Master-Scene language established
š Start with Establishing Shot
š Closer shots to see subjects, action
š Close-ups to see emotion, reaction
š Wider shots for context, reinforcement
š Return to Closer shots to reinforce reaction
š Often Wide end shots to show resolution, exit
™ ‘The Line’ gains acceptance as standard
The Visual Systems Lab

1910-1920: Languages Develop


™ Parallel Language established
š Action in multiple locations occurring
simultaneously – ‘cross-cutting’
™ Experimentation with convention
š Colour used to indicate time, emotion
™ New technology enables moving cameras
™ Optical transitions become common
The Visual Systems Lab

The Lonedale Operator (1911)

Clip of the cross-cut sequence of


the woman telegraphing for help
as the robbers try to break in
The Visual Systems Lab

1920-1930: Conceptual Language


™ Pudovkin tests ‘Constructive’ model
š Meaning of film can be broader than
content of individual shots
š Enables conceptual communication
š Creates tension, enhances action
™ Eisenstein expands this through ‘Montage’
š Graphic, Rhythmic, Picture-Sound and
Ideological juxtaposition
The Visual Systems Lab

Battleship Potemkin (1925)

Clip of the Odessa Steps sequence


The Visual Systems Lab

1930-onward: Grammar Explored


™ Visual Discontinuity
š Buñuel, surrealists challenge perception and
expectation
š Exploit audience understanding of continuity
š Continuity, Lead Space, Eye Line ‘abused’
š Development of Graphic Match cuts
š Advent of Jump Cuts to jar viewer
™ If squeamish, look away now!
The Visual Systems Lab

Un Chien Andalou (1929)

Clip of the graphic match between


the cloud crossing the moon and the
lady’s eye being sliced
The Visual Systems Lab

1930-onward: Grammar Explored


™ Transition Driven
š Focus audience attention explicitly
š No attempt to hide technique
š Most common in newsreels
š Continues in ‘hard sell’ advertising
The Visual Systems Lab

‘News on the March’ (1941)

Clip of the ‘news reel’ from


Citizen Kane
The Visual Systems Lab

1930-onward: Grammar Explored


™ Dissection of Action (Vorkapich)
š Similar to montage but salient shots shown
to convey a linear process
š Focuses audience on steps involved, detail
š Widely used today in conjunction with other
techniques
The Visual Systems Lab

Cool Hand Luke (1967)

Clip of the opening sequence where


Luke cuts the heads off parking meters
The Visual Systems Lab

1930-onward: Evolution of Sound


™ Initially based on Theatre model
š Music accompaniment to enhance drama
™ Evolved to basic recording of event
š Dialogue, action but all live
™ Multi-track recording, editing enabled control
š Foley, looping, better synchronisation
™ Sound design evolves to control how
picture is interpreted
The Visual Systems Lab

Modern Filmmaking
™ Draws on all previous languages
š Exploits audience knowledge and expectation
of grammar and genre convention
™ There is little that is new…
š Re-invention, re-discovery of technique
™ … except what technology enables
š Snorkel lens, virtual cameras, etc.
The Visual Systems Lab

Raging Bull (1980)

Clip of the Sugar Ray Robinson


title fight that Lamotta loses
The Visual Systems Lab

Automated Parsing
™ Can technique and language be of use?
š Picture and Sound analysis can reveal
patterns to potentially classify language
š Parsing certain elements could enable higher-
level semantic concepts to be identified
š Determination of grammar could better
define search ranges
š The goal is reduce the scale of the domain,
splitting it into manageable data sets
The Visual Systems Lab

Picture Analysis
™ Basic technical attributes can be readily
parsed through image processing
š Pans, Tilts, Zooms and transitions can be
found using disparity measures
™ Camera movement can be classified
based on 3D extraction from images
š Dolly moves, Crane shots, Steadicam, etc
The Visual Systems Lab

Picture Analysis
™ Relative motion and subject tracking can
be extracted via segmentation
™ Face Detection methods can facilitate
the extraction of Framings and Groupings
š Noting orientation of respective head positions
can indicate Eye Line
™ Face Recognition can aid with context
™ Text recognition through OCR to add detail
The Visual Systems Lab

Picture Analysis
™ Object pattern matching techniques can
be applied in stages
š Location or Setting can be established based
on landmark detection (Eiffel Tower = Paris)
š More abstract settings can be detected by
salient features (traffic lights = street)
™ The goal is to build semantic database
to enhance chances of correct parsing
The Visual Systems Lab

Sound Analysis
™ Segmentation of soundtrack into music,
dialogue and effects for individual parsing
™ Classification based on sound signatures
noting time to correlate with picture
š Give clues as to what to search for visually
š Enhance accuracy of parsing through
reinforcement
The Visual Systems Lab

Film Language Analysis


™ Patterns of picture and sound characteristics
in shots can provide insight into language
š Framings and durations can create signature
š Types of Continuity can be classified
™ Could be used to determine genre
™ Could be used to assist dating of programs
™ Can provide additional detail for
summarisation
The Visual Systems Lab

Summarisation
™ Rich Metadata and prosaic textual
descriptions can be generated
™ Potential cross links with related programs
to confirm/enhance extracted data
™ Benefit to visual summarisation unclear
š Dependent on metaphor used
š Static depictions can nullify effect of editing
technique
The Visual Systems Lab

Summarisation
™ Reverse Storyboarding Metaphor
š Based on film preproduction visualisation
methods
š Draws on extracted technical attributes
of pictures
š Automatically adds visual cues in keeping
with storyboarding technique
• onion skins, streaks, trail lines, arrows, field cuts, etc.
š Annotated, mosaic or combination form
The Visual Systems Lab

Reverse Storyboarding

Tilt Up, Pan Left in Arrow Form Tilt Up Reference


The Visual Systems Lab

Reverse Storyboarding

Zoom Out in Mosaic Form Zoom In Reference


The Visual Systems Lab

Issues and Obstacles


™ Film language very rich and not uniform
™ Patterns of editing not always consistent
™ Significant knowledge of film history
required to create effective model
™ Accurate training and testing difficult given
diversity of programs in existence
™ Computationally very expensive
The Visual Systems Lab

Conclusions
™ Film languages have evolved into a
sophisticated and multi-faceted grammar
™ Techniques that facilitate these languages
are constantly growing and changing
™ Understanding language and technique
could enable better semantic extraction
™ Textual Summarisation could benefit but
relevance to visual summaries unclear
The Visual Systems Lab

Useful References
Technique of Film and Video Editing
Ken Dancyger, Focal Press, 3rd ed., 2002

A Director’s Method for Film and TV


Ron Richards, Focal Press, 1992

Film Directing Shot by Shot


Steven Katz, Michael Wiese Prods, 1991
The Visual Systems Lab

Thank You!

John Mateer
Dept of Electronics
Dept of Theatre, Film & TV
(011 44) 1904 433 245
jwm10@ohm.york.ac.uk

You might also like