News & Notices

Analog Devices
JPEG2000 Image Codec

From May 1999 - May 2002, I worked with Analog Devices Video Division in the development of their ADVJP2000 image codec. This is the first (and to-date the only - June 2002) hardware implementation of the JPEG2000 specification in existence. This chip is mentioned in Taubman & Marcellins book, JPEG2000: Image Compression Fundamentals, Standards, and Practice, as well as being the cover story for issues of both EETimes and the May 2001 issue of Advanced Imaging Magazine.

During the development process many experiments were performed in efforts to better understand some of the more subtle aspects of the specification, some of which are posted below. This is the work of both Alex Zatsman, the leading software developer for the JPEG2000 team at Analog Devices, and myself.
Rate / Distortion Experiments
Tiling Artifact Experiments
Image Regions of Interest
MPEG2 -Video clip of an image decoded for incremental bitrates
Optimal Truncation of Arithmetically Coded Data -- How many bits do you save?
JPEG2000 Coding of Video Sequences -- Which filter is better for video 9/7 or 5/3?

Mallat Diagrams and the Wavelet Transform
The Mallat decomposition is a common method of viewing various levels of an image wavelet transform. Here the hierarchical nature of the transform as specified in the specification is represented geometrically by rectangular regions of the image domain. The Mallat decomposition images of the graylevel version of Jimi Hendrix is shown proceeding from left to right with 1 and 2 levels of wavlet transforms respectively.

Dr. Andrew R. Willis

Associate Professor

University of North Carolina - Charlotte
Electrical and Computer Engineering Dept.
9201 University City Boulevard
Charlotte, NC 28223-0001


Phone  (704) 687-8420
Fax    (704) 687-4762


My interests lie in computer vision, pattern recognition, medical and natural image processing, and the general field of processing discrete geometries in multiple dimensions. Specific problems on which I currently work are the measurement and processing of 2D range images, 3D surface scans, and 4D data sets created by successive 3D surface scans over time.



Research results and progress is submitted here.