Page 138 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience
P. 138

Section 4.6.  Block-Matching Methods                          115



                        Foreman @ 8.33 f.p.s.
               36
                   With overlapping
                   No overlapping
               34
              PSNR Y  (dB)   32
               30
               28

               26
               24
                1  5   10  15  20  25  30  35  40  45  50
                            Frame
                                                          th
                      (a) Prediction quality     (b) Original 15 frame at 8.33 f.p.s














              (c) Compensated using BMA (29.35 dB)  (d) Compensated using OMC (30.93 dB)

                            Figure 4.9:  Comparison  between OMC and BMA


            4.6.7  Properties of Block-Motion Fields and Error Surfaces

            This subsection presents some basic properties of the BMME algorithm when
            applied  to  typical  video  sequences.  These  properties  will  be  utilized  and  ref-
            erenced  in  subsequent  chapters  of  the  book.  All  illustrations  in  this  subsec-
            tion were generated using a full-pel full-search block-matching algorithm with
            16 × 16  blocks,  ±15  pels  maximum  displacement,  restricted  motion  vectors,
            SAD as  the BDM, and original  reference  frames.

            Property 4.6.7.1  The  distribution  of  the  block  motion   eld  is  center-biased.
            This means that smaller displacements are more probable and the motion vector
            (0; 0)  has  the  highest  probability  of  occurrence.  In  other  words,  most  blocks
            are stationary or quasi-stationary. This property is illustrated in Figure 4.10(a)
            for  AKIYO  at  30 frames=s  (frame  skip  of  1).  The  property  also  holds  true  for
   133   134   135   136   137   138   139   140   141   142   143