CAREER: Fast Efficient Adaptive Filter Banks and Applications in Digital Multimedia Coding/Processing

Johns Hopkins University, Baltimore MD

Investigators

Abstract

CAREER: Fast Efficient Adaptive Filter Banks and Applications in Digital Multimedia Coding/Processing Tran Project Summary With the recent emergence in popularity of the Internet, the demands for and the interests in digital multimedia are growing exponentially. A high-performance filter bank is typically at the heart of every state-of-the-art digital multimedia system. For example, current international image/video compression standards JPEG, MPEG, H.261 and H.263 are all based on the 8x8 Discrete Cosine Transform (DCT), an 8-channel 8-tap orthogonal filter bank. Of great importance is the ability to design a filter bank that can fully exploit (i) the statistical properties of a particular signal or class of signals; (ii) the goals of the applications; (iii) the computational resources available at the processing systems. Particularly, fast, efficient, and low-cost transforms have played a crucial role in revolutionary advances throughout the history of digital signal processing. This research involves the development of next-generation fast efficient adaptive filter banks for applications in digital multimedia coding/processing. The implementations are VLSI-friendly, entirely integer-based, possibly even multiplier-free, while still maintaining a high level of coding performance. Moreover, all of the filter banks are constructed from modular cascades of similar building blocks where the same basic components are employed repeatedly in a highly regular fashion, resulting in smaller chip area and lower power consumption. The investigator incorporates as many desirable transform properties onto the structure as possible, leading to a much smaller space of free parameters and hence improving the effectiveness of optimization algorithms, especially in real-time adaptive systems. Well-proven techniques in optimization, linear prediction, and adaptive filtering theory are being applied toward the lattice structure optimization. The new family of filter banks is expected to provide versatility, flexibility, and adaptivity in time-frequency mapping, coding performance, cost of implementation as well as integration into modern digital multimedia processing/communication systems. ------------------------------------------------------------------- Level of Effort Statement. At the recommended level of support, the PI will make every attempt to meet the original scope and level of effort of the project. Trac D. Tran