International Journal of Multimedia and Ubiquitous Engineering
Vol.10, No.7 (2015), pp.295-304
http://dx.doi.org/10.14257/ijmue.2015.10.7.31
ISSN: 1975-0080 IJMUE
Copyright ⓒ 2015 SERSC
All Phase Biorthogonal Transform Based on GPU
Rongyang Shan, Chengyou Wang
*
, Xiao Zhou and Liping Wang
School of Mechanical, Electrical and Information Engineering, Shandong
University, Weihai 264209, China
sdusry@163.com, wangchengyou@sdu.edu.cn, zhouxiao@sdu.edu.cn,
sduwlp@163.com
Abstract
In this paper, all phase biorthogonal transform (APBT) based on parallel algorithm is
proposed. It can solve two problems. First, block-based DCT transform coding has
serious blocking artifacts when the image is highly compressed at low bit rate. Second,
APBT can solve the problem about blocking artifacts, but it does not have a fast
algorithm, it has a low efficiency when APBT applies to image processing. So APBT
based on parallel algorithm can solve the above problems, and it provides more space for
improving the processing speed of APBT. We use the CUDA toolkit based on GPU which
is released by NVIDIA to design the parallel algorithm of APBT. Experimental results
show that the maximum speedup ratio of parallel algorithm of APBT can reach more than
40 times with a very low version GPU, compared with conventional serial APBT. And the
reconstructed image using the proposed algorithm has the same performance with the
serial one in terms of objective quality and subjective effect. The proposed parallel
algorithm based on GPU of APBT also can be used in image compression, video
compression, the edge detection, and some other fields of image processing.
Keywords: GPU; Parallel Computing; All Phase Biorthogonal Transform (APBT);
Discrete Cosine Transform (DCT)
1. Introduction
At present, the research for the discrete cosine transform (DCT) has been developed in-
depth. Discrete cosine transform (DCT) [1] is widely used in the field of image processing.
DCT is applied to the standards of the international image compression and video
compression, like JPEG [2], MPEG-2 [3], MPEG-4 [4], H.264/AVC [5] and
H.265/HEVC [6]. Although discrete wavelet transform (DWT) took the place of discrete
cosine transform in JPEG 2000, DCT still takes an important place in image processing.
With the development of orthogonal transform, DCT has become quite mature, and two-
dimensional DCT is the core of JPEG coding. However, DCT is not the best choice in
image coding, because block DCT transform coding has serious blocking artifacts when
the image is highly compressed at low bit rate. The all phase biorthogonal transform
(APBT) [7] which is based on Walsh-Hadamard transform (WHT), DCT and inverse
discrete cosine transform (IDCT) proposed by Hou et al. is a new transform for image
compression instead of DCT, which solves the problem of blocking artifacts in DCT, and
APBT uses the uniform quantization step instead of the complex quantization table in
DCT which makes APBT save the storage space of quantization Table.
Currently, parallel computation based on GPU is more and more popular in scientific
computation, because GPU has more cores than CPU, and the parallel computation based
on GPU can bring higher efficiency in complex computation than CPU. CUDA toolkit
which is released by NVIDIA makes parallel computation based on GPU easier than
before. Parallel algorithm can get 10 times acceleration easily than serial algorithm by