The Eyeball AnyBandwidth Technology is the revolutionary adaptive video
technology developed at Eyeball Networks for real-time one-way and two-way
streaming audio-video applications. It enables audio-video communications
among multiple clients and/or servers having heterogeneous processor
capacities and network connections, and/or network connections with
dynamically changing properties. AnyBandwidth provides optimized video
quality (such as frame rate, picture resolution, and sharpness) to each user
based on the available network bandwidth and processing power.
Eyeball AnyBandwidth™ Technology supports a wide variety of
hosts and network connections. It dynamically adapts voice and video quality
of each client to the actual networking conditions and available hardware
resources.
The key feature of the
AnyBandwidth Technology is that the video quality is individually optimized
for each client according to the available resources at the client's end.
Thus clients connecting via T1 lines, cable modems, or DSL technology, can
decode high quality video at high frame rate, and, at the same time, clients
connecting via 56k dial-up-modems or wireless connections send or receive
motion video at lower quality and lower frame rate. For example, when
resources are abundant, clients receive full-motion and full-quality video
(e.g., 352x288 pixels at 30 frames/sec). However, in the case of limited
network bandwidth and/or CPU power, clients receive gracefully and optimally
reduced video (e.g., 88x72 pixels at 7.5 frames/sec) without experiencing
any blackouts, large delays, or blocking artifacts.
Current Technologies
Traditionally, video streaming solutions are based on the
so-called switching technology. In this system, the video source provides
multiple copies of the same video at different frame rates and quality
levels. Each copy is independently transmitted and decompressed as shown in
Figure 2.
Figure 2: Traditional switched video streaming system.
The major disadvantages of such a system include:
It requires several compressors to run in parallel, resulting in large CPU
overhead.
If there are multiple receivers with different networking and hardware
capabilities, multiple streams, each at a different frame rate and/or
quality, must be transmitted independently resulting in large network
overhead.
Adaptation to network conditions and hardware resources must be
source-driven, i.e., based on some feedback from the client. The video
source must select the stream to be transmitted. This scheme works well for
one-to-one communications but is clearly not applicable for many-to-many
communications.
Highly Scalable Encoding
Traditional video compression standards such as MPEG-2 or
H.263 are not suitable for scalable video applications. On the other hand,
the AnyBandwidth Technology is based on patent-pending technology providing
scalable video representation.
Eyeball AnyBandwidth provides frame rate and picture quality scalability
(resolution and sharpness) that is designed to support both network
bandwidth availability and processing capacity adaptation. This technology
can be used to support a wide variety of clients with CPUs ranging from
Pentium 166 MHz MMX to Pentium 4 1.5 GHz and beyond, and network bandwidth
ranging from 10 kbps to 1 Mbps. Eyeball AnyBandwidth™ is the only video
technology in the industry that can support such a wide range of processors
and network connections.
The block diagram of Eyeball AnyBandwidth scalable video compression system
is shown in Figure 3. The scalable compressor produces multiple video
layers. During transmission, each video layer is transmitted in a separate
substream. The video quality at the receiver depends on the number of video
layers used in the decompression. Decompressing more video layers results in
better video quality.
Figure 3: Eyeball AnyBandwidth™ scalable video streaming.
The advantages of the AnyBandwidth Technology include:
-
No CPU power is wasted since only one compressor is
required.
-
Aggregate output bandwidth is optimized by using
non-overlapping sub-streams; clients only receive video layers that can be
supported by their network connections and processing hardware.
-
Adaptation to network conditions and hardware resources is
client-driven, i.e., each client decides how many video layers to receive
and process. This scheme works well with both one-to-one and many-to-many
communications.
The AnyBandwidth Compression Engine: How It
Works
The block diagram of the Eyeball
AnyBandwidth compression engine is shown in Figure 4. The
developed technique efficiently exploits the temporal
redundancy inherent in video by determining parts of the
frame to be compressed. Then, block-based discrete cosine
transform (DCT) is applied followed by our novel data
reorganization technique. A subset of DCT coefficients is
selected, which has significant contribution to the video
quality. As seen in Figure 4, only a small fraction of DCT
coefficients is significant and needs to be transmitted to
the client resulting in very high compression ratio. During
compression, the significance information is efficiently
stripped among multiple video layers.
Figure 4: Eyeball AnyBandwidth™ compression engine.
The block diagram of the AnyBandwidth decompression engine
is shown in Figure 5. The decoder receives a subset of video
layers based on the network connection and available
processor capacity, and performs the reverse operation of
the compression engine.
Figure 5: Eyeball AnyBandwidth™ decompression engine.
Eyeball AnyBandwidth Compression Performance
The performance of the AnyBandwidth Technology is
illustrated for typical video communications scenarios.
Figure 6 illustrates the adaptation capability of
AnyBandwidth Technology ranging from bit rate of 10 kbps at
2.5 frames/sec to bit rate of 200 kbps at 10 frames/sec.
Figure 6: Better video quality requires more video layers to
be received demanding higher bandwidth.
Visual comparison of decompressed images is shown in Figure
7. The uncompressed video would require 3 Mbps, which easily
overwhelms even high-speed networks. With the AnyBandwidth
Technology, bit rates ranging between 10 kbps and 200 kbps
and frame rates ranging between 2.5 frames/sec and 10
frames/sec can be obtained by partially decompressing only
one video stream.
Figure 7: Visual performance comparison with
the AnyBandwidth Technology. (a) Original image. (b)
Decompressed image at 10 frames/sec, 200 kbps. (c)
Decompressed image at 5 frames/sec, 68 kbps. (d)
Decompressed image at 2.5 frames/sec, 10 kbps.
