Evolution Technologies Home

Welcome to Evolution Technologies - Home of AdaptMedia (TM)

AdaptMedia is a network video stream player that adapts to its environment in order to give the best quality of video possible. Just like our ancestors had to adapt to changing environments, so does AdaptMedia. If the network is heavily congested, AdaptMedia recognizes this, and changes the configuration of its internal decoding engine to make the best use of the available bandwidth. The result is improved image quality maintained across diverse changing conditions.

AdaptMedia consists of 2 core components:

AdaptNet: recognizes and predicts changing network behavior
AdaptVideo: uses network information to optimize compression parameters

AdaptMedia Product highlights:

AdaptMedia is fully portable and has been tested on the following platforms:
- Win95, Win98
- WinNT
- Linux
AdaptMedia enhances displayed image quality by using linear interpolation

AdaptMedia Technical Information

The objective of Adaptmedia is to implement a cross platform network adaptive video encoder, decoder and player. AdaptMedia encodes video in two layers:

Base layer: highly compresses the video quality by using a quantization step size 31.
Enhancement layer: gives an added layer of resolution encoding the difference between the base layer and actual video frame, using a quantization step size of 4.

The following components have been implemented:

Video player for yuv image file format
Video coding/decoding with H.263 standard support
Video network layer for live transmission
PSNR scalability with base and enhancement layer

AdaptMedia Video Player

AdaptMedia video player 1.0 has just been released. Here is a look at the player:

AdaptMedia video player with linear interpolation

AdaptMedia Video Streaming Encoder Decoder

AdaptMedia's video streaming supports PSNR scalability at two levels, base and enhacement mode. If the network is congested, AdaptMedia will tradeoff resolution to decrease network use and avoid dropping frames.

Experiment setup

AdaptMedia video streaming capabilities have been tested using a video stream server machine and a decoding and player machine. The machines used were 266Mhz Intel Pentium II with 64Mb of RAM and 10 Mbps ethernet running WinNT.
The experiment was setup using the trevor video sequence with 150 frames, encoded as qcif of size 5.43Mb. The encode bit stream sizes are as following:

Using no scalability and quantization step size of 4, size = 203 Kb
Using scalability:
- Base layer at quantization step size of 31, size = 41 Kb
- Enhacement layer at quantization step size of 4, size = 119 Kb
- Total size (base + enhacement) = 160 Kb

AdaptMedia's network adaptive performance was tested by setting up a video stream channel and in addition, congesting the network with 10 equivalent transmiter receiver pairs on the server. All of the tramisters were run on the server machine as given above. The recievers used consisted of an actual decoder and 10 'dummy' decoders, all being run on the same decoder machine as given above. These 'dummy' decoders all use the same network adaptive algorithm as the actual decoder, the difference being that they don't actually decode the bitstreams in order to avoid overloading the machine.
Each server transmits video stream at the following rates:

Using no scalability: 2.1 Mbits/sec
Using PSNR scalability: 1.6 Mbits/sec

Therefore, we can see that given these bitrates, 10 decoders would congest a 10 Mbps connection.

Experimetal Results

Given the experiment setup described above, the following percentage packet drop were observed:

Using no scalability: 27.4 %
Using PSNR scalability: 17.05 %

Therefore we can see that using AdaptMedia's network adaptive video streaming, less packets are dropped. This results in less number of dropped frames due to the decoder being unable to decode the whole the whole frame due to missing bits.
The perceived video quality for each of the two cases, measured from 1-5, where 1 is worst performance, was observed to be:

Experiment	1	2	3	4	5
AdaptMedia Scalable	3	3	3	3	4
Mon-Scalable	2	3	1	1	3

Link to our code

AdaptMedia Team

Last modified: Thu May 6 15:54:48 EDT 1999