Graphics Reference
In-Depth Information
H.264/AVC obtains the quarter-sample values by first obtaining the values of
nearest half-pixel samples and averaging those with the nearest integer samples,
according the position of the quarter-pixel [
27
]. However, HEVC obtains the
quarter-pixel samples without using such cascaded steps but by instead directly
applying a 7 or 8-tap filter on the integer pixels.
In H.264/AVC, a bi-predictively coded block is calculated by averaging two uni-
predicted blocks. If interpolation is performed to obtain the samples of the uni-
prediction blocks, those samples are shifted and clipped to input bit-depth after
interpolation, prior to averaging. On the other hand, HEVC keeps the samples
of each one of the uni-prediction blocks at a higher accuracy and only performs
rounding to input bit-depth at the final stage, improving the coding efficiency by
reducing the rounding error.
The details of these features are presented in the following sections.
5.3.1.1
Redesigned Filters
An important parameter for interpolation filters is the number of filter taps as it has a
direct influence on both coding efficiency and implementation complexity. In terms
of implementation, it not only has an impact on the arithmetic operations but also
on the memory bandwidth required to access the reference samples. Although the
6-tap filter for estimating half-pixel positions in H.264/AVC produces a constant
phase delay of 0.5 for all frequency components due to symmetry, the passband
(range of frequencies where the magnitudes are relatively unaltered) is not large.
Increasing the number of taps can yield filters that produce desired response
for a larger range of frequencies which can help to predict the corresponding
frequencies in the samples to be coded. Considering modern computing capabilities,
the performance of many MCP filters were evaluated in the context of HEVC and a
coding efficiency/complexity trade-off was targeted during the standardization.
Consider the design of a half-pixel interpolation filter with 2N taps denoted
as
h
D
Œh
0
;h
1
;
;h
2N 1
T
. Due to the desired half-pixel symmetry only
N coefficients can be different, which can be denoted in the form
h
D
Œh
0
;h
1
;
;h
N 1
;
;h
1
;h
0
T
. Now consider the interpolation of a DC signal
(with all samples equal). It is desired that the interpolated value be the same as
the input. Hence, we require 2
P
N 1
n
0
h
n
D
1, also known as the normalization
constraint. This further reduces the number of degrees of freedom from N to N
1.
In the case of a quarter-pixel interpolation filter however, only the normalization
D
condition
P
2N 1
n
0
h
n
D
1 is imposed as symmetry is not necessary, which gives
2N
1 degrees of freedom. The aim of interpolation filter design is to determine
these degrees of freedom so as to remain close to the desired frequency response.
Here a brief overview of the design of HEVC interpolation filters is provided.
For a detailed explanation the reader is referred to [
17
]. The basic idea is to forward
transform the known integer samples to the DCT domain and inverse transform
D
