Implementing the 2-D Wavelet Transform on SIMD-Enhanced General-Purpose Processors

Abstract

Abstract-The 2-D Discrete Wavelet Transform (DWT) consumes up to 68% of the JPEG2000 encoding time. In this paper, we develop efficient implementations of this important kernel on general-purpose processors (GPPs), in particular the Pentium 4 (P4). Efficient implementations of the 2-D DWT on the P4 must address three issues. First, the P4 suffers from a problem known as 64K aliasing, which can degrade performance by an order of magnitude. We propose two techniques to avoid 64K aliasing which improve performance by a factor of up to 4.20. Second, a straightforward implementation of vertical filtering incurs many cache misses. Cache performance can be improved by applying loop interchange, but there will still be many conflict misses if the filter length exceeds the cache associativity. Two methods are proposed to reduce the number of conflict misses which provide an additional performance improvement of up to 1.24. To show that these methods are general, results for the P3 and Opteron are also provided. Third, efficient implementations of the 2-D DWT must exploit the SIMD instructions supported by most GPPs, including the P4, and we present MMX and SSE implementations of horizontal and vertical filtering which provide a maximum speedup of 3.39 and 6.72, respectively.

Optimizing Gaussian filtering of volumetric data using SSE

“…Further, SIMD optimizations are often used in signal (1D) and image (2D) processing. In [24], SIMD was used to efficient implement the 2D wavelet transform. SIMD vectorization of histogram functions is described in [25].…”

A robust wavelet based feature extraction method for face recognition

“…Giving a multiresolution analysis in both time and frequency domain and having different alternatives for the basis function makes wavelet transform a better candidate than Fourier transform [4,5] for many applications including JPEG2000 [6]. A growing number of publications that deal with hardware implementation of wavelet transform [7][8][9][10] are another proof of its applicability.…”

A novel way of face recogniton to improve the quality of features extraction

“…Wavelet transform [13] has been widely used in many fields such as JPEG2000 [14].A growing number of publications deal with hardware implementation of this transform [15][16][17][18] which demonstrates its utility in the area of Pattern Recognition. A multi-resolution analysis of a signal with localization in both time and frequency is the advantage of wavelet transform over Fourier and cosine transforms [19][20][21]. In order to perform feature extraction, the two-dimensional (2-D) wavelet transform of the faces is calculated using different basis functions only up to 1 scale .…”