An Efficient Mode Decision Algorithm for Combined Scalable Video Coding - Advanced Concepts for Intelligent Vision Systems - page 58

(CABAC), etc [1]. Especially, mode decision including spatial prediction for in-
tra mode and variable block size motion estimation/compensation (ME/MC)
with multiple reference frames for inter mode is much more complicated for
SVC because the inter layer prediction is used between layers. Therefore, it is
necessary to design a method to reduce this complexity with a minimal loss of
image quality.
Many kinds of fast mode decision schemes have been proposed for H.264/AVC:
Fast variable block size motion estimation (ME) [3], fast inter coding mode se-
lection [4][5][6], fast intra prediction [7], etc. Recently, some fast mode decision
algorithms for SVC have been reported. However, almost every fast algorithm
for SVC has been specialized in a single scalability. A fast mode decision al-
gorithm for spatial scalability has been suggested by Li et al. [8] in which the
mode distribution relationship between the base layer and enhancement layers
is used. Lim et al. [9] proposed a fast encoding mode decision method using
an early skip mode detection technique based on the relationship between the
temporal levels in a group of pictures (GOP). Some literatures try to combine
scalability [10][11]. A fast mode decision algorithm by Li et al. [10] can support
partially combined scalability: spatial scalability, a coarse grain signal-to-noise
ratio (CGS), and temporal scalability. They use the correlation of mode distri-
bution between the base layer and enhancement layers. A layer adaptive mode
decision algorithm and a motion search scheme by Lin et al. [11] have been
proposed for CGS and temporal scalability in which modes with limited con-
tributions to the coding eciency are skipped based on a statistical analysis in
order to reduce the computational complexity of the mode search.
In this paper, we propose a fast mode determination scheme for inter frame
coding supporting temporal, spatial, and quality scalability based on correlative
information between the base layer and enhancement layers. In the proposed
algorithm, we define a cost function for the motion area based on ordered mode
information and two mode search classes: large block type (16 × 16, 16 × 8, 8 × 16)
and sub block type (8 × 8, 8 × 4, 4 × 8, 4 × 4). Based on the designed cost function
for the motion area, either large block type or sub block type is assigned to the
mode search process. Next, we determine the direction of mode search (forward,
backward, bidirectional) using the direction of first mode type for the determined
block type. Using this direction, mode search for the remaining modes in the
blocktypeisperformedtofindthebestmodetypeforthecurrentmacroblock
(MB).

2 Inter Frame Coding for Combined SVC

SVC supports three special types of scalability that allow complete images to be
decoded from only part of the bitstream. The three types are spatial, temporal,
and quality. In addition, each scalability can be combined to support general
condition. Because SVC is an extension of H.264/AVC, most advanced coding
techniques for H.264/AVC are used for inter frame coding. In this section, we
briefly introduce the type of scalability and the mode decision process.