Block-Based Approach to 2-D Wavelet Transform on GPUs

Abstract

This paper introduces a new approach to computation of 2-D discrete wavelet transform on modern GPUs. The proposed approach involves block-based processing enabling one seamless transform even for high resolution input data. Inside the blocks, two distinct methods can be used -either separable or non-separable 2-D lifting scheme. Furthermore, the paper presents a comparison of the proposed approach under different conditions to the best existing methods, whereas our approach consistently outperforms the other ones. Our methods are implemented using the OpenCL framework and tested on a wide range GPUs.

Parallel wavelet schemes for images

“…Since this work is based on our previous work in [23], it should be explained what the difference between this work and [23] is. In [23], we presented a block-based method employing a scheme foregoing the schemes proposed in this paper.…”

“…The block- based approaches in various forms were also used in, e.g., [5; 14; 19; 23]. Since the block-based approaches overcome the row-column ones (as shown in [23], or analyzed in [19]), we do not include the classical row-column methods in our performance comparison. Instead, we only compared different schemes employed under the block-based approach.…”

“…In [23], we presented a block-based method employing a scheme foregoing the schemes proposed in this paper. Unlike [23], the schemes presented in this paper are defined by general predict and update operators.…”

Parallel Wavelet Schemes for Images

“…Since this work is based on our previous work in [23], it should be explained what the difference between this work and [23] is. In [23], we presented a block-based method employing a scheme foregoing the schemes proposed in this paper. Unlike [23], the schemes presented in this paper are defined by general predict and update operators.…”

“…based approaches in various forms were also used in, e.g., [5; 14; 19; 23]. Since the block-based approaches overcome the row-column ones (as shown in [23], or analyzed in [19]), we do not include the classical row-column methods in our performance comparison. Instead, we only compared different schemes employed under the block-based approach.…”