Transcript 下載/瀏覽Download

An Efficient Search Algorithm for Motion
Estimation
Shung-chih Chen, Tzyy-kuen Tien, and Chung-wei Tsai
ABSTRACT
An efficient search algorithm using the principle of Hexagon-Based Search
algorithm (HEXBS) for motion estimation called Hexagon-Based Fast Search
algorithm (HEXFS) is proposed in this paper. The proposed algorithm combines the
hexagon search algorithm and a two-step search algorithm which uses cross search
patterns to reduce the huge calculation complexity. The experimental results show
that the search points’ improvement ratio of the modified search algorithm can be
higher than other conventional algorithms which have been proposed.
PROPOSED ALGORITHM
-15
-10
-5
0
5
10
15
-15
-10
-5
0
5
10
15
Step 1.1
Step 2.1
Step 1.2
Step 2.2
Step 3
Step 4
Fig. 1 Search example leading to the motion vector (-4, -10)
for our HEXFS algorithm.
Fig. 1 shows an example of our HEXFS
algorithm. First of all, we search around
the center point by Step 1.1 and Step 1.2
respectively. In Step 1.1, if the result
meets the early termination criterion, we
go to Step 4, but if not, go to Step1.2. In
Step 1.2, if the minimum block
distortion point is found at the center, go
to Step 3; otherwise, we go to the Step
2. We use XCSP on Step 2.1. After
finding the best coarse search point, we
use RCSP to refine it in Step 2.2, and
then we use large hexagon-based search
pattern on Step 3 and small hexagonbased search pattern on Step 4 to get the
best fine search point.
EXPERIMENTAL RESULTS
Algorithm
File
Full
Search
Diamond Hexagon
Proposed
TSCDS
Search
search
HEXFS
TSS
4SS
SHIBA (640x480)
208.752
24.063
25.829
31.291
25.886
26.055
22.414
Baseball (1024x768)
211.265
24.210
26.010
31.523
26.050
26.154
22.493
Diner (1600x1200)
212.773
24.293
26.110
31.652
26.141
26.249
22.544
Average
210.93
24.189
25.983
31.489
26.026
26.153
22.483
* No. of search points per macro block