Research Document Imaging Hybrid Processing Hybrid Compression

Search

Image Compression
in Hybrid Raster/Vector Storage Systems
Dr. Eugene Ageenko

Why hybrid? MOTIVATIONS.

• Over 9,000,000 engineering drawings in the world (1997)
• Only 15 % are in computer format
• The rest MUST BE digitized
• Suitable FORMAT for storage?
• VECTOR (CAD/CAM) format
  1. easy to modify and re-scale images but
  2. expensive to convert and not of high quality
  3. is really necessary?
  4. only in parametric modelling and automated control systems
     (< 15 % of existed applications)
• RASTER format is mainly sufficient
  1. images are well compressible by JBIG
  2. storage size is about the same as for vector file
  3. vectorizing only by demand or partially

Engineering drawing (Example)

HYBRID RASTER-VECTOR FORMAT

• consist of RASTER and VECTOR representation of IMAGE
• RASTER image – array of pixels:
• exact reconstruction of image in all conditions
• further recognition of desired parts (eg. OCR) if can not be done using existing available technologies
• VECTOR image – list of vector primitives:
• benefit of vector editing
• suitable for semantic analysis

PROBLEM: store raster and vector files separately? No!

GOALS

• reduce the file size compared both to full vectorizing and to raster image compression
• allow lossless reconstruction of the image in all conditions
• support further processing of the image by the vectorizing

HYBRID storage system diagram (© 1998 Eugene Ageenko, Pasi Fränti)

• local spatial dependencies exploited by context-modelling
• global dependencies – by vectorizing
• reconstructed (feature) image is used to improve prediction accuracy

Compression with JBIG using 2-level context template:

Example

HYBRID File Structure

VECTOR DATA

1. end points
2. width

RASTER DATA

1. JBIG-style compressed original
2. vector features utilized in model
3. optional morphological filtering of mismatch between original and reconstructed images