What is Auto-Focusing?

  • The sharpness of an image is the determining factor in human visual perception of focus.  Upon examining the frequency components of an unfocused image, you will find that there are relatively few high frequency components. As the image comes into focus, high frequency components increase.
  • In analog cameras, auto-focusing is done manually, with a sensor which actually measures distance (object distance) to the photographed object.  In digital cameras, as mentioned in the Motivation section, signal processing (computational power) may be leveraged to enable auto-focusing.
  • The lens of a camera is essentially  a low pass filter, which focuses by shifting the cutoff frequency of the filter to the higher frequency side.  In digital cameras, the software controls the lens position, maximizing higher frequency components.  The term focus measure refers to the metric by which we measure the amount of high frequency components--maximum focus measure occurs when the image is most closely focused.

3D Shape Recovery

  • Basic of 3D shape recovery from image focus analysis

    -  the objects at different distances in front of a lens are focused      at different distances behind the lens

  • Two types of techniques based on the image focus analysis for 3D shape recovery.

        - Shape From Focus (SFF) is a search method which searches the             camera  parameters (lens position and/or focal length) that             correspond to focusing the object. SFF uses a sequence of images             taken by changing the focus setting of the imaging optics in small             steps. For each pixel, the focus setting that maximizes image contrast             is determined. This, in turn, can be used to compute the depth of the             corresponding scene point.

        - Shape From Defocus (SFD), unlike, SFF is not a search method. It             measures the amount of blur in an image and needs as low as only two             images with different optical settings to obtain the depth map of an             entire scene. It is faster than SFF but less accurate. The relative             defocus in the two images can, in principle, be used to determine 3D             structure. The focus level in the two images can be varied by changing             the focus setting of the lens, or by changing the aperture size.


  Digital Watermarking
  Intelligent Digital Satellite Receiver System
  PCB Soldering Inspection
  Digital Signal Processing
  Image Processing
  Machine/Robot Vision
  Statistical Signal Processing
  Pattern Recognition
  Image Compression


Digital Video


Visual Communications