Stroke Extraction Offline handwriting recognition Noise, faint - - PowerPoint PPT Presentation

stroke extraction
SMART_READER_LITE
LIVE PREVIEW

Stroke Extraction Offline handwriting recognition Noise, faint - - PowerPoint PPT Presentation

Stroke Extraction Offline handwriting recognition Noise, faint strokes, and gaps Related Work Skeletonization Local Search Graph Search Intelligent Scissors Least-cost search of an image to find an edge Weighted


slide-1
SLIDE 1

Stroke Extraction

  • Offline handwriting

recognition

  • Noise, faint strokes,

and gaps

slide-2
SLIDE 2

Related Work

  • Skeletonization
  • Local Search
  • Graph Search
slide-3
SLIDE 3

Intelligent Scissors

  • Least-cost search of an image to

find an edge

  • Weighted sum of several gradient

features

  • Optimal path from seed point to

every other point in the image

slide-4
SLIDE 4

Intelligent Pen - Kevin Bauer

  • Like Intelligent Scissors but

without user interaction

  • Multiple expanding wavefronts
  • Creates new seeds on the fly as

it builds out

slide-5
SLIDE 5
slide-6
SLIDE 6
slide-7
SLIDE 7

Intelligent Pen

slide-8
SLIDE 8
slide-9
SLIDE 9
slide-10
SLIDE 10
slide-11
SLIDE 11
slide-12
SLIDE 12
  • Think “edge detection on trees”
  • Find the difference between the

wavefront counts for two nodes

  • Keep track of largest difference
  • Threshold on differences

Finite Differences

50 100 10 160 160 160 150 50 100 Wavefront

slide-13
SLIDE 13
slide-14
SLIDE 14
slide-15
SLIDE 15

Path Cooling / New Seed

  • Large wavefront are slow and

error prone

  • Instead, smaller wavefront
  • Create new seeds at end points
slide-16
SLIDE 16
slide-17
SLIDE 17
slide-18
SLIDE 18
slide-19
SLIDE 19

Future Work - Include Direction

slide-20
SLIDE 20
slide-21
SLIDE 21

Thank You