Optical Recognition - Features

• Pre-treatment and pre-classification
  - Skew correction
  - Resizing
  - Detection of position of staves
  - Cleaning of the image
  - Pre-classification of elements (frames and borders, ornate letters, lyrics, inter-staff elements and titles)
  
• Staff recognition using HMM
  - Extraction function with implicit treatment of staff lines
  - Treatment of staff curvature
  - Beam search decoder (adapted from Torch)
  - Integration and ponderation of music model
• Dynamic optimization using maximum a posteriori (MAP) adaptation
  
• Creation of music models
  - Witten-Bell discounting
  - Katz backoff-ing
  - Decomposition of music model
  
• Graphic interface for viewing of original image and results with synchronized view (zoom and scroll)
  
• Integrated music editor for correction of any recognition errors (adapted from Wolfgang)
  
• Superimposition of results on to original for better detection of any recognition errors
  
• Batch processing (pre-treatment, recognition and model creation)
  
• Exporting of results (Wolfgang format)
  
• Multilingual