Video Significance-Linked Connected Component Analysis

MCVL Homepage Department of Computer Engineering and Computer Science

Contact Information:

Xinhua Zhuang
Department of Computer Engineering and Computer Science
University of Missouri-Columbia
Columbia, MO 65211

Jozsef Vass
Department of Computer Engineering and Computer Science
University of Missouri-Columbia
Columbia, MO 65211


Table of Contents


Video Significance-Linked Connected Component Analysis

Video significance-linked connected component analysis (VSLCCA) is a very advanced video compression scheme developed at the Department of Computer Engineering and Computer Science at the University of Missouri-Columbia by Jozsef Vass, Bing-Bing Chai, and Xinhua Zhuang. It is based on our still image compression algorithm termed significance-linked connected component analysis (SLCCA) developed by Bing-Bing Chai, Jozsef Vass, and Xinhua Zhuang.


Comparison with H.263+

Performance comparison is done on the first 300 frames (10 seconds) of the following sequences at 10 frames-per-second (100 frames, use only every third frame) in QCIF (144x176) resolution. For H.263+, the source code is downloaded form Signal Processing & Multimedia Group of the University of British Columbia. In the comparison, we use the Akiyo, Foreman, and Silent Voice sequences (download from our ftp site).


Implementation

The development of VSLCCA was done in C language. Later, the algorithm was implemented in JAVA by Jozsef Vass and John Harwell. To increase decoding speed, the algorithm was been implemented as a Netscape helper (standalone application).

[VSLCCA Decoder]

The main purpose of the file menu to load and save a new stream. It is possible to load a stream from the local disk, or from a remote location. In the letter case, the HTTP address of the stream should be specified.

[File Menu]

There are few options that are available through options menu. The sequence can be displayed in doubled size, and can be played repeatedly. The sequence can also be stopped and the frame can be captured as a PPM image.

[Options Menu]


Demonstration

The MPEG-4 standard test image sequence "Akiyo" is used as the first demonstration. The original sequence contains 300 frames in QCIF resolution (144 rows with each row having 176 pixels). The time duration of the sequence is 10 seconds which results in 30 frames-per-second frame rate.

Download the original frames. Each of the 50 frames is 76032 bytes (144*176*3, where 144 is the number of rows, 176 is the number of pixels in each row and each frame has a red, green, and blue channel). When you are using a 28.8 kbps modem, to download the whole sequence might take approximately as long as 18 minutes. The display is a simple JAVA interface. To download the original sequence, click here

Here the performance of MPEG-1 and VSLCCA is compared. Two examples are given. In the first one, the frame rate is 5 frames-per-second (fps), and the required bit rate is 10k bit-per-second (bps). Thus the compression ratio is 152:1. In the second example, the frame rate is increased to 10 fps and the bit rate is increased to 24 kbps. This results in 129:1 compression ratio.

5 fps, 10 kbps MPEG-1 VSLCCA JAVA VSLCCA
10 fps, 24 kbps MPEG-1 VSLCCA JAVA VSLCCA

The VSLCCA algorithm is also very well suited for remote sensing data compression.


Code Availability

The native VSLCCA decoder is now available only for SGI workstations. Now, we are porting the code for SUN, and HP workstations. For code, please contact John Harwell or Jozsef Vass


Trouble Shooting

The algorithm was tested on several platforms. The JAVA version requires Netscape Communicator version 4.00 or above.

The JAVA version have been tested on the followings platforms:

The native version is currently running on SGI workstations (O2 running IRIX 6.3 and Octane running IRIX 6.4).


Feedback

Please mail comments, suggestions, and bugs to John Harwell or Jozsef Vass


[Table of Contents]    [Home]CECS Multimedia Communications and Visualization Laboratory