V-Lab: Visual Media

V-Lab is conducting research in computer vision for next generation communication and entertainment applications. The primary research challenge is the introduction of methodologies to understand complex natural scenes from images.

Visual media encompasses research in visual content production, visual interaction and visual content analysis from 3D content production through to video annotation for search and retrieval of large scale media archives.

Visual Content Production

In visual content production our goal is to bridge-the-gap between real and computer generated imagery. Research combines the fields of computer vision, graphics and animation to develop new methods for reconstruction, modelling and understanding of the real-world from images and video. Our core expertise is in the area of multiple view video analysis for 3D reconstruction, modelling and animation. Applications include: sports analysis (soccer, rugby, athletics), 3D TV and film production, visual effects, character animation for games, and digital doubles for both film production and visual communication. V-Lab is recognised as a leading research group in the area of 3D Video processing and 3D human shape and motion analysis from video. Recent research advances include technologies for stereo 3D production from monocular cameras, free-viewpoint video in sports and high-realism interactive characters from actor performance capture.

Interactive characters from actor performance capture

Interactive characters from actor performance capture

Visual Search and Retrieval

The growing volume of digital visual media shared online motivates new techniques for the search and management of image and video collections.  CVSSP has established an internationally leading track record in the metadata annotation and visual search of image and video collections.  Projects include object and activity recognition, sketch based visual search, pose search in video, and photo based visual search for detection of illegal or plagiarised content.

Visual Stylization

Research in visual stylization seeks to develop new tools for artistically stylized rendering of visual media – to emulate artistic effect or provide new creative tools for digital manipulation of content.  Projects include the transformation of photos into sketches, drawings and paintings – and of video into cartoons.  Current work includes the artistic rendering of consumer media collections to create abstract compositions from home videos, and learning to paint portraits to imitate the style of artists.  This Computer Graphics has motivated fundamental research in Computer Vision, e.g. in temporally coherent video segmentation and scene understanding.

Visual Interaction and Communication

Research in visual communication includes the development of methods for  understanding human movement  and non-verbal communication in facial expression, gestures and body language. This research is developing novel interfaces for human computer interaction together with modelling, understanding  and synthesis of human behaviour. Research in facial analysis has introduced technologies for video-rate 3D face capture and tracking to analyse and synthesise the dynamics of facial movement during expression and speech. V-Lab has introduced techniques for acquisition of 3D Video of facial shape and appearance. Subsequent 3D facial tracking provides the basis for both analysis of facial dynamics and modeling to create highly realistic digital doubles. The goal of this research is to introduce natural animation and communication interfaces for understanding and synthesis of human body and facial movement.

V-Lab has established strong collaborations and technology transfer with UK companies in the creative industries including broadcast (BBC,Sony BPRL) film post-production (Framestore, DNeg, The Foundry, FilmLight) and interactive entertainment (Sony, EA, Rebellion, Lionhead, Kuju).

Projects

V-lab is active in a portfolio of research projects supported by EPSRC, EU, TSB, industry  and the Royal Society. The research activity is underpinned by an EPSRC platform grant in Visual Media providing strategic long-term funding for adventure and continuity of research. Projects include:

  • Visual Media Platform Grant (EPSRC)
  • RE@CT Immersive Production and Delivery of Interactive 3D Content (EU IST) 
  • SCENE Novel Scene Representations for Richer Networked Media (EU IST) 
  • SyMMM Synchronising Multimodal Movie Metadata (TSB)
  • BodyShape Body Shape Recognition for Online Fashion (EPSRC)
  • Royal Society Industry Fellowship with Framestore Digital Doubles: from real-actors to computer generated digital doubles.

Previous Projects

People

 

Free-viewpoint video in sports

Free-viewpoint video in sports

 

 

Digital doubles for audio-visual speech animation

Digital doubles for audio-visual speech animation

 

 

High-resolution 3D video face capture

High-resolution 3D video face capture

 

 

Non-linear video browser developed as part of VIDI-Video (EU FP6)

Non-linear video browser developed as part of VIDI-Video (EU FP6)

 

Artificially generated portrait oil painting

Artificially generated portrait oil painting

 

Coherent video segmentation for scene understanding

Coherent video segmentation for scene understanding

Page Owner: jc0028
Page Created: Wednesday 5 October 2011 14:05:26 by lb0014
Last Modified: Thursday 26 May 2016 15:02:09 by pg0016
Expiry Date: Saturday 5 January 2013 14:04:36
Assembly date: Wed Aug 24 19:58:05 BST 2016
Content ID: 65945
Revision: 8
Community: 1379

Rhythmyx folder: //Sites/surrey.ac.uk/CVSSP/research activity/labs/v-lab: visual media
Content type: rx:Generic