23-06-2012, 04:50 PM
HIGH COMPRESSION OF FACES IN VIDEO SEQUENCES FOR MULTIMEDIA APPLICATIONS
[attachment=25315]
INTRODUCTION
Image and video coding are one of the most
important topics in multimedia processing and
communications. During the last thirty years we have
witnessed a tremendous explosion in research and
applications in the visual communications field.
However, and in spite of all this effort, there are some
applications that still demand higher compression
ratios than those provided by state of the art
technologies.
In particular, and due to its high
applicability, there is a need to provide novel
compression schemes to encode faces present in
video sequences. Although the new standards H.263+
[1] and the synthetic part of MPEG-4 [2] along with
other model-based proposed schemes [3] achieve
high compression ratios for this particular
application, we still believe that further compression
is needed, among others, for mobile and video
streaming environments.
IMAGE CODING THROUGH RECOGNITION
Introduction
Many proposals have been made in the last years for
image and video coding. In particular H.263+ is
mainly intended for low to high data rate robust
compression and is based on a block-based
redundancy removal scheme [1]. In addition, MPEG-
4 combines frame-based and segmentation-based
approaches along with model-based video coding in
the facial animation part of the standard which allows
efficient coding as well as content access and
manipulation [2]. It can be said that H.263+ and
MPEG-4 represent the state of the art in video coding
[6].
Face coding using a Principal Component Analysis approach
Let us simplify the visual content by assuming that
we are interested in the coding of faces in a video
sequence. Let us also assume that automatic tools to
detect a face in a video sequence are available. Then,
some experiments show that a face can be well
represented by very few coefficients found through
the projection of the face on an eigenspace previously
defined [4]. The image face can be well reconstructed
(decoded), up to a certain quality, by coding only
very few coefficients.
FIXED EIGENSPACE APPROACH
In order to check the validity of the eigenspace
approach for image coding, results using a fixed
eigenspace will be first presented. These results will
be useful to point out the main drawbacks of the
eigensapce approach and to fully understand the
adaptive eigenspace proposed in the next section.
ADAPTIVE EIGENSPACE APPROACH
Introduction
In this section we propose an eigen coding approach
that adapts itself to the face content appearing in the
video sequence. Notice that a related technique has
been presented in [9] although the approach is
significantly different.
The initial encoding scheme follows that of
the fixed eigenspace explained in Section 3. In that
scheme, any important change in the expression of
the face will lead at a poor performance of the
scheme. In order to overcome this problem, we have
designed a fall back mode system, which consists of a
quality of reconstruction evaluation block followed
by an upgrade mechanism of the coder and decoder
eigenspace databases.
[attachment=25315]
INTRODUCTION
Image and video coding are one of the most
important topics in multimedia processing and
communications. During the last thirty years we have
witnessed a tremendous explosion in research and
applications in the visual communications field.
However, and in spite of all this effort, there are some
applications that still demand higher compression
ratios than those provided by state of the art
technologies.
In particular, and due to its high
applicability, there is a need to provide novel
compression schemes to encode faces present in
video sequences. Although the new standards H.263+
[1] and the synthetic part of MPEG-4 [2] along with
other model-based proposed schemes [3] achieve
high compression ratios for this particular
application, we still believe that further compression
is needed, among others, for mobile and video
streaming environments.
IMAGE CODING THROUGH RECOGNITION
Introduction
Many proposals have been made in the last years for
image and video coding. In particular H.263+ is
mainly intended for low to high data rate robust
compression and is based on a block-based
redundancy removal scheme [1]. In addition, MPEG-
4 combines frame-based and segmentation-based
approaches along with model-based video coding in
the facial animation part of the standard which allows
efficient coding as well as content access and
manipulation [2]. It can be said that H.263+ and
MPEG-4 represent the state of the art in video coding
[6].
Face coding using a Principal Component Analysis approach
Let us simplify the visual content by assuming that
we are interested in the coding of faces in a video
sequence. Let us also assume that automatic tools to
detect a face in a video sequence are available. Then,
some experiments show that a face can be well
represented by very few coefficients found through
the projection of the face on an eigenspace previously
defined [4]. The image face can be well reconstructed
(decoded), up to a certain quality, by coding only
very few coefficients.
FIXED EIGENSPACE APPROACH
In order to check the validity of the eigenspace
approach for image coding, results using a fixed
eigenspace will be first presented. These results will
be useful to point out the main drawbacks of the
eigensapce approach and to fully understand the
adaptive eigenspace proposed in the next section.
ADAPTIVE EIGENSPACE APPROACH
Introduction
In this section we propose an eigen coding approach
that adapts itself to the face content appearing in the
video sequence. Notice that a related technique has
been presented in [9] although the approach is
significantly different.
The initial encoding scheme follows that of
the fixed eigenspace explained in Section 3. In that
scheme, any important change in the expression of
the face will lead at a poor performance of the
scheme. In order to overcome this problem, we have
designed a fall back mode system, which consists of a
quality of reconstruction evaluation block followed
by an upgrade mechanism of the coder and decoder
eigenspace databases.