Sven Grottke

Apriori Rate Allocation in Wavelet-Based Image Compression

We present a rate allocation scheme that pre-computes optimal quantization bucket sizes based on a mathematical model of wavelet transformed natural images prior entropy encoding. We combine our scheme with the JPEG2000 embedded rate-allocator and find in our experiments that it is able to increase JPEG2000 encoding speed by a factor of two without losing image quality.

Scalability of a distributed virtual environment based on a structured peer-to-peer architecture

We investigate the scalability of distributed virtual environments (DVEs) based on a structured peer-to-peer (P2P) overlay. We focus on network load and message routing latency. To this end, we study a prototypical DVE consisting of a simple game scenario and a P2P architecture based on Pastry and Scribe as proposed by Knutsson et al. [1]. Both our theoretical analysis and simulation results show that under constant population density, the network load is constant except for the overhead messages incurred by the overlay protocol. The overall network load per host as well as the message routing latency grows logarithmically with the number of hosts; this is in partial contradiction to Knutsson et al.s results. We propose a resolution to this contradiction.

Adam A Testbed for Distributed Virtual Environments

In distributed virtual environments (DVEs) the data on which the hosts operate is not consistent at all times. To restore data consistency, the DVE has to employ a consistency algorithm. Unfortunately, all existing DVEs have been built for specific application scenarios, which makes it impossible to compare the consistency algorithms and to choose a suitable candidate for a new scenario. To overcome this, we have created a modular simulator-based DVE testbed named Adam with the ability to plug in different application scenarios as well as different consistency algorithms and network constraints. The testbed also contains a large set of measurement tools. Our testbed currently supports two application scenarios and several of the most common consistency algorithms found in the literature. We can compare the solutions on an objective scale and confirm that optimistic consistency typically outperforms loose consistency.

Dealing with Multiple Truths in Online Virtual Worlds

Virtual environments and online games are becoming a major market force. At the same time, the virtual property contained in these environments is being traded for real money and thus attains a real value. Although the legal issues involved with this virtual property have not yet been decided, they will have to be soon. It is foreseeable that the next generation of very large virtual worlds will carry the possibility of multiple truths existing at the same time. Under such circumstances, it will be impossible to physically protect virtual property. In order to protect virtual property, virtual environment systems will therefore have to conform to certain requirements. We analyze what these requirements are in order to either prevent cheating or at least prove a digital offence has transpired. Along with greater security, this will also simplify end user support, which is one of the major cost factors for online games. [Article copies are available for purchase from InfoSci-on-Demand.com]

An Integrated Course on Wavelet-Based Image Compression - Learning Abstract Information Theory on Visual Data

We describe the implementation of and our experiences with a capstone course on wavelet based image compression held at the Berlin University of Technology in the years 2002 to 2006. This course has been designed as an “integrated project”, which means that it combines lectures, seminar talks to be prepared and held by the students, and a programming part. The design goal of this course has been to provide all the necessary theoretical knowledge to understand the concepts behind image compression technologies, such as JPEG2000. We are also aiming at simulating the work-flow as found within an IT company as realistically as possible, preparing electrical engineers and computer scientists as well as possible for their professional life. This training does not only include the technical, but also the social skills required to successfully complete larger projects. The subject of image compression offers the advantage of requiring a solid knowledge on terms of information science such as entropy, distortion, quantization, Fourier and wavelet-transformation, but also offering a direct visual feedback of how these techniques perform. Therefore, we believe that image compression is an attractive topic to be used for a capstone course. Traditionally, a course would assign weekly programming exercises to the students; however, we believe this to be unsuitable for a capstone course as it does not simulate the work-flow of a professional software development team; furthermore, it does not require the degree of team-work we deem critical to modern software development. Thus, we divide students into groups of two to four people and assign each team to one sub-task of an image codec and provide some boiler-plate code of our own. Much to their astonishment, students soon find themselves spending a considerable amount of time with project management and coordination activities. That means, teams have to design interfaces and data structures to combine their efforts to create a working project, which adds an often underestimated social component to the course. With some guidance from the teachers, students have always been able to supply a working code at the end of the semester. Needless to say, the thrill of having a nontrivial working program at the end of the course is a major source of motivation for our students and adds much to the satisfaction and positive feedback we receive.

Apriorirate Allocation in Wavelet-Based Image Compression

We present a rate allocation scheme that pre-computes optimal quantization bucket sizes based on a mathematical model of wavelet transformed natural images prior entropy encoding. We combine our scheme with the JPEG2000 embedded rate-allocator and find in our experiments that it is able to increase JPEG2000 encoding speed by a factor of two without losing image quality.

High-bitrate approximation

We are reviewing scalar quantizers with deadzone and overload in the high-bitrate approximation, give sharp bounds on the quality of this approximation and present results on the optimal quantizer in this setting. Some recent results by Hui and Neuhoff are reproduced under weaker conditions. We conclude by comparing the mathematical results with experimental data. For details about the results of this article we refer to an article in prepeparation.