Resource-driven transmission, display, and processing of multimedia in mobile devices

Yamini Nimmagadda, Purdue University


There has been a dramatic growth in the usage of rich media such as images and videos in mobile devices for fields such as personal communication, education and entertainment, and surveillance. The mobile devices are resource-constrained; they have limited energy, small displays, limited memory, and low processor speeds. These resource-constraints pose great challenges for processing, display, transmission and reception, and usage of multimedia in mobile devices. This dissertation provides solutions for (1) energy conservation, (2) content adaptation, and (3) faster image processing in mobile devices. For (1) energy conservation: transmission of media consumes significant amounts of energy. This energy is proportional to the media size. This dissertation presents rendering techniques to reduce the media sizes by removing fine details. The techniques select rendering parameters based on wireless bandwidth and media content to achieve positive energy savings. The techniques are implemented on an HP iPAQ PDA; 20% and 24% net energy savings are observed on average for images and videos respectively. For (2) content adaptation: the mobile devices have small displays. The content should be adapted to fit on the displays. This dissertation presents adaptation techniques for presentations constituting media with different start times and durations. The adaptation is based on preferences and temporal constraints specified by content providers and layouts are automatically generated by computing locations and start times. This dissertation compares three solutions to generate layouts: exhaustive search, dynamic programming, and greedy algorithms. For (3) faster image processing: Mobile robots are used for computation intensive applications such as moving object recognition and tracking. Existing robots have low-end embedded processors with clock speeds in the range of few hundred MHz. This dissertation presents a computation-offloading framework to divide the tasks between the robot and a high-performance server to reduce the response time of the robot. Our method reduces the execution time by up to 92% and tracks an object that moves five times faster.




Lu, Purdue University.

Subject Area

Computer Engineering|Electrical engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server