An Efficient Mode Decision Algorithm for Combined Scalable Video Coding - Advanced Concepts for Intelligent Vision Systems - page 61

Macroblock Partition Modes

16 × 16 16 × 8 8 × 16 P8 × 8

2 6 7

1 4 5

3 8 9

Sub-Macroblock Partition Modes

8 × 8 8 × 4 4 × 8 4 × 4

S2 S6 S7

S1 S4 S5

S3 S8 S9

Fig. 4. Inter Prediction Modes

where s and r ( MV ) mean the current block and the predicted block with the
estimated MV , respectively. SSD is the sum of squared differences

between the current block s and its corresponding block r . λ motion is the Lagrange
multiplier and R ( MVD,REF ) means the bit rates for encoding the motion
vector difference MVD and the number of reference frame REF .

{ s, r ( MV ) }

Intra Mode Prediction. In addition to the inter MB coding types, various
intra prediction modes are specified in H.264/AVC as shown in Fig. 5. Unlike
the previous coding standards, intra prediction in H.264/AVC is performed in
the pixel domain by referring to neighboring samples. First, prediction blocks
are formed by directional interpolation using the intra prediction modes. Next,
we calculate the RD cost of the difference between the current MB and its
corresponding prediction block. The best intra mode is determined as one having
the smallest RD cost. The RD cost function is

J intra = SSD { s, r } + λ mode · R ( s, r, M ) , (2)

where λ mode is the Lagrange multiplier and R ( s, r, M ) means the bit rates for
encoding the residual data according to the predicted mode M .

The Best Coding Mode Selection. To obtain the best MB coding mode for
the P-frame, H.264/AVC encoder exhaustively tests all possible encoding modes
including inter modes and intra modes. As a result, the mode having the smallest
RD cost is determined the best coding MB mode among all possible modes in
the P-frame.

The Problem of Mode Decision for Combined SVC. A simple structure
for combined SVC with inter layer prediction is shown in Fig. 3. When the inter
layer prediction is off, the multi-layer signals are equal to multiple independent
sequences transmission. In order to improve the encoding eciency, the inter