High-Speed Transform Coding Architectures for Video Communications

Abstract

The field of signal processing has developed dramatically over the last several decades owing to applications in such diverse fileds as speech, image, and video communication, biomedical engineering, acoustics, consumer electronics, and many others. Discrete sinusoidal transforms such as discrete cosine transform (DCT), discrete sine transform (DST), and discrete Hartley transform (DHT), discrete Fourier transform (DFT), Lopped Orthogonal Transform (LOT), and Complex Lapped Transform (CLT) are powerful tools in many applications of signal processing. Due to the advances in ISDN network and high definition television (HDTV) technology, real-time processing of speech, image and video signals become very desirable. Clearly the high computational rates required in HDTV systems cannot be achieved by general-purpose parallel computers because of severe system overheads. The only way to meet the high computational rates of real-time signal processing is by developing special-purpose architectures which exploit the regularity, recursiveness, and locality of the signal processing algorithms. We focus on developing real time VLSI algorithms and architectures for discrete sinusoidal transforms for video applications.