Bilinear algorithms for discrete cosine transforms of prime lengths

Authors:
Venkatram Muddhasani;Meghanad D. Wagh
Affiliations:
Department of Electrical and Computer Engineering, Lehigh University, Bethlehem, PA;Department of Electrical and Computer Engineering, Lehigh University, Bethlehem, PA
Venue:
Signal Processing - Signal processing in UWB communications
Year:
2006

Citing 5
Cited 3

Discrete cosine transform: algorithms, advantages, applications

Discrete cosine transform: algorithms, advantages, applications
A new fast algorithm for computing prime-length DCT through cyclic convolutions

Signal Processing
Face Recognition Using the Discrete Cosine Transform

International Journal of Computer Vision - Special issue: Research at McGill University
DCT algorithms for composite sequence lengths

IEEE Transactions on Signal Processing
New fast recursive algorithms for the computation of discretecosine and sine transforms

IEEE Transactions on Signal Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a strategy to design bilinear discrete cosine transform (DCT) algorithms of prime lengths. We show that by using multiplicative groups of integers, one can identify and arrange the computation as a pair of convolutions. When the DCT length p is such that (p - 1)/2 is odd, the computation uses two (p - 1)/2 point cyclic convolutions. When (p - 1)/2 = 2m q with m 0 and q odd, the computation requires one (p - 1)/2 point cyclic convolution and a combination of a q point cyclic convolution and a 2m point Hankel product. Using bilinear algorithms for convolutions and Hankel products, one gets a bilinear DCT algorithm. We also show that the additions required beyond the convolutions can be minimized by a small modification to the convolution algorithms. This minimization exploits the fact that efficient bilinear convolution algorithms are almost always based on Chinese Remainder Theorem.