Docstoc

Motion The research report Block matching algorithm Motion

Document Sample
Motion The research report Block matching algorithm Motion Powered By Docstoc
					The research report
    Block matching algorithm
    Motion compensation
    Spatial transformation




                      Xiaomei Yu


                               Jump to first page
          Motion compensation:
   Motion compensation methods can be defined as
    techniques that divide images into local regions (block or
    patches) and estimate a set of motion paramaters for
    each region.
   The procedure is to synthesizes the predicted image of
    the nth frame        from the decoded image of the
    previous frame          can be regarded as an image
    warping or texture mapping process. It can be written
    as
    where the geometic relationship between         and
    is defined by the transformation funcations       and
    .


                                                 Jump to first page
      Motion compensation:
   Block matching algorithm (BMA) estimates the
    motion of blockes and transmits the estimated
    motion vectors. BMA can be regarded as a
    method that applies spaial transformations on
    the blocks of the image.
   Pel-recursive algorithm (PRA) estimates the
    motion of each pixel and does not transmit the
    motion information.




                                       Jump to first page
       Block matching algorithm
   BMA adopts a motionmodel that describes the motion
    of the image objects by the translational motion of
    blocks. The problems of BMA (e.g. blocking artifacts)
    are caused by the insufficiency of the motion model to
    represent the real world. To overcome this defect,
    many coding methods have been proposed.
   BMA can be regarded as a method that applies spatial
    transformations on the blocks of the image. By
    generalizing this transformation, new motion
    compensation methods that adopt different motion
    models can be developed.



                                             Jump to first page
   For example in BMA, the transformation
    functions for the pixels in the i th block of
    the image are



   Various motion compensation methods can
    be composed by adopting different
    transformation functions.
    Transformation function:
    ----- Approximate the original motion vector field woth a
    smaller number of parameters.
    ----- A smaller amount of computation.



                                                         Jump to first page
   The total amount of computation in texture mapping is a
    sum of the following three processes: (a) computation
    of the motion parameters (b) computation of the
    transformation function ( c) interpolation of intensity
    values. The later two are important since they are
    performed each pixel. The computational cost of (b)
    depend on the existence of an effective scan line
    algorithm, which scans the image computing f(x, y) and
    g(x, y) for each pixel.




                                                Jump to first page
Spatial transformation
   Affine transformation (AFMC)



   Bilinear transformation (BLMC)



   Perspective transformation



                                     Jump to first page
       The advantages/disadvantages
   The parameters can be determined from motion vectors of
    three vertices. Then, the patches of this transformation are
    triangles. The cost in computing the motion parameters
    remains cheap regardless of the shape of the patch. An
    effective scan line algorithm exists.
   Only rectangular patches are allowable since otherwise the
    computation of the motion parameters become
    complicated. An effective scan line algorithm with two
    additions per pixel exists for this transformation.
   The scan line algorithm requires two divisions for each
    pixel.


                                                   Jump to first page

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:3
posted:2/23/2012
language:
pages:8