Digital Media Research Group Computer Science Department University of Bristol Bristol, UK

1
Digital Media Research GroupComputer Science
DepartmentUniversity of BristolBristol, UK
Areas of Interest

• Computer Vision, Image Processing, Speech,
• Graphics and Animation

CV/IP People
Prof. Barry Thomas Dr. Majid Mirmehdi Dr. Neill
Campbell Dr. Andrew Calway 16 RAs and PhD
Students
2
Research Focus

• Image and video segmentation
• Motion analysis and tracking
• 3-D from 2-D images and video
• Industrial Inspection
• Medical Image Analysis
• and more

3
Locating text
Five statistical measures in small
neighbourhoods across the image

• Histogram variance
• Edge count
• Continuous spatial density
• Symmetrical edge distribution
• Spread of edge angles

NN
Classification by a neural network
30
4
Fronto-parallel Recovery
05
5
Fronto-parallel Recovery

03
6
Generating higher resolution images with an
active camera

• Zoom in and pan over text region
• High resolution image reconstructed from mosaic

7
Tracking In Vitro Groundtruth and In vivo
Patella Tendon
Shortened Trajectory History
8
Tile Abnormality Detection
Post Processing
Gabor Filters
WHT Method
9
Texture Synthesis
10
Automatic Identification of Exudative Retinopathy

• Use computer vision techniques to diagnose
  diabetic eye disease
• Identify exudates in images of the retina by
  image segmentation

Exudates
Typical retinal image
Normalisation
Contrast enhancement
Exudates
Colour segmentation
Segmentation
11
Locating the Optic Disk

• Locating the optic disk is important for further
  analysis
• Use mathematical morphology and snakes to
  localise the disk

Optic disk
Blood vessels removed
Final snake location
Ground truth
Retinal image
Localised boundary
Hand-drawn boundary
Comparison
12
Visual Surveillance and Understanding Human
Behaviour

• Analysing CCTV surveillance data to detect
  behaviour and action
• Predict the onset of anti-social behaviour
• Raise the alarm
• Bring CCTV channel to operators attention
• zoom in and take a close up image