Configurable image data shaping algorithms and architectures for wireless multimedia communication.

摘要

With the introduction of higher capacity wireless networks, as well as richer wireless handsets, there is a growing demand for image-based mobile services, from MMS messages in the consumer space to critical applications in medicine and security. However, a major challenge in enabling wireless image-based data services is the need to support the desired quality of service under varying network and server load conditions. To address these technical challenges, this dissertation proposes robust and cost-efficient image adaptation algorithms and architectures.; First, for enabling quality-aware, resource-informed shaping of images, we develop a new Energy/Latency-efficient Wavelet-based Image Tuning (E/LeWIT) method. The developed E/LeWIT identifies image-compression parameters in JPEG2000 which can be used for fine-grained image tuning. We also propose a runtime reconfiguration algorithm to perform fast image adaptation by tuning the above mentioned parameters.; We next address the architectural issues associated with enabling adaptive image data shaping through a novel configurable hardware/software architecture. The proposed configurable hardware/software architecture provides ASIC-like performance improvement while retaining the flexibility necessary to implement the dynamic image adaptation and to accommodate different image compression algorithms.; We further extend the usability of the proposed image shaping algorithms by demonstrating the efficiency either for adapting to the time varying wireless network conditions (network-aware dynamic image adaptation) or to the resource constrained service load variations at the servers ( server-aware image quality adaptation). While each of the dedicated image adaptation techniques can provide significant performance improvement by adapting separately to each of the service variability issues, the overall service variability existing at the total end-to-end wireless image data service environments affects the total quality of service in a combined way. In order to provide guaranteed end-to-end quality of service, we finally propose a "dynamic end-to-end image adaptation" technique. The proposed image adaptation technique can indeed allow the system to deliver a high quality of image content, along with a guaranteed quality of service level.