• Object Extraction
  • Background subtraction is a method for extracting moving objects. It is used in many fields in computer vision because it is implemented easily and the process of it can be handled very fast. However, some problems exist. Objects under illumination changes cannot be extracted by background subtraction. We propose a robust method to illumination changes.
 
 
 
 
 
  • Object Tracking
  • Object tracking by particle filter attracts attention because it is robust to noises and changing conditions. It sometimes fails in tracking when a target object intersects with other objects with similar patterns. We propose a method for robust tracking to the intersection with similar objects.
 
 
 
 
 
 
 
 
  • Shadow Detection
  • Many methods for object extraction have a problem that shadows are extracted as moving objects. Shadows often have a harmful effect on the result. We propose a method for shadow detection.
 
 
  • GPGPU
  • Graphics Processing Unit (GPU) is very powerful. It has been used not only for standard graphics operations but also for other processes. We study about speeding up by GPU. A GPU based approach for object extraction is proposed.

  • Shape from Images
  • We have been studying about recovering 3D shape from some images. Recent researches include 3D modeling and neural network based rendering, shape recovery from an image or defect classification for electronic board.