A high-throughput, memory efficient architecture for computing the tile-based 2D discrete wavelet transform for the JPEG2000

Dimitroulakos, Gregory; Galanis, Michalis D.; Milidonis, A.; Goutis, C.E.

doi:10.1016/j.vlsi.2004.11.002

Cited by 8 publications

(15 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The low-pass output coefficient is recursively passed through similar filter banks for higher resolution levels [8]. For one decomposition level, the two filter outputs y 0 [n] and y 1 [n] are expressed according to the following equations [1]:…”

Section: Brief Presentationmentioning

confidence: 99%

“…The h[n] and w[n] are the high-pass and low-pass filters, respectively, while the y 1 [n] and the y 0 [n] are the high-pass and the low-pass sub-band coefficients [4] (denoted hereafter H and L), respectively. After the filtering operation, the produced sequences are downsampled by a factor of 2 [1,4].…”

Section: Brief Presentationmentioning

confidence: 99%

“…After the filtering operation, the produced sequences are downsampled by a factor of 2 [1,4]. Figure 1.…”

Section: Brief Presentationmentioning

confidence: 99%

“…The advantage of DWT is that it carries out a multiresolution analysis of a signal with localizations in both time and frequency. The DWT is a highly efficient and flexible technique for sub-band decomposition of signals [1]. A lot of architectures that implement the 2D-forward (2D-FDWT) and the inverse (2D-IDWT) have been presented [1].…”

Section: Introductionmentioning

confidence: 99%

“…The DWT is a highly efficient and flexible technique for sub-band decomposition of signals [1]. A lot of architectures that implement the 2D-forward (2D-FDWT) and the inverse (2D-IDWT) have been presented [1]. These architectures are composed of filters 662 C. SOUANI, I. GAZZAH AND K. BESBES for the 1D-transformation associated with memory units for storing the coefficient results, which are re-injected into filters [1].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Design and implementation of optimized architecture for computing the 2D‐DWT for JPEG2000 compression

Souani

Gazzah

Besbes

2009

Circuit Theory & Apps

View full text Add to dashboard Cite

SUMMARYTo obtain an effective and efficient system, a multitude of architectural solutions must be assessed and their performances compared. Design and evaluation of several hardware architectures require enormous time for their development and the evaluation. In this paper, an exploration of possible architectures for the 2D-discrete wavelet transform (DWT) is accomplished. Particularly, filters used by the standard JPEG2000 are studied: the lossy 9/7 and the lossless 5/3 filters. The designs have been coded with high description level using Handel-C language and the target hardware is Xilinx FPGA of the Virtex-4 family. It is known that the technique of in-place calculation of the 2D-DWT by lifting scheme allows saving space memory, but the problem is that the computed coefficients are not stored in consecutive addresses, and a system for address decoding becomes necessary. The address decoding must be made at every new decomposition level. We propose then a new simple and efficient technique that allows storing the calculated coefficients in a homogeneous space memory and in consecutive addresses. No additional memory or address decoding is necessary. Finally, the designed system allows a rather high throughput and optimized number of hardware resources.

show abstract

Section: Brief Presentationmentioning

confidence: 99%