23-02-2013, 12:13 PM
A High-Performance VLSI Architecture for Image Compression Technique Using 2-D DWT
[attachment=51591]
Abstract
Image compression is the application of Data compression on digital images. A fundamental shift in the image compression approach came after the Discrete Wavelet Transform (DWT) became popular. To overcome the inefficiencies in the JPEG standard and serve emerging areas of mobile and Internet communications, the new JPEG2000 standard has been developed based on the principles of DWT. In this research, an architecture that performs both forward and inverse lifting-based discrete wavelet transform is proposed. The proposed architecture reduces the hardware requirement by exploiting the redundancy in the arithmetic operation involved in DWT computation. The proposed architecture does not require any extra memory to store intermediate results. Using Verilog HDL, the encoder for the image compression employing DWT was implemented. This architecture has been described in VHDL at the RTL level and simulated successfully using ModelSim simulation environment. In this paper, high-efficient lifting-based architectures for the 5/3 discrete wavelet transform (DWT) are proposed. The proposed parallel and pipelined architecture consists of a horizontal filter (HF) and a vertical filter (VF). The system delays of the proposed architectures are reduced. Filter coefficients of the biorthogonal 5/3 wavelet low-pass filter are quantized before implementation in the high-speed computation hardware. In the proposed architecture, all multiplications are performed using less shifts and additions. The proposed architecture is 100% hardware utilization and ultra low-power. The architecture has regular structure, simple control flow, high throughput and high scalability. Thus, it is very suitable for new generation image compression systems, such as JPEG- 2000. Keywords: 5/3 discrete wavelet transform (DWT), IDWT, horizontal filter (HF), vertical filter (VF), lifting-based architecture, JPEG-2000. 1. INTRODUCTION Data compression is the technique to reduce the redundancies in data representation in order to decrease data storage requirements and hence communication costs. Reducing the storage requirement is equivalent to increasing the capacity of the storage medium and hence communication bandwidth. Thus the development of efficient compression techniques will continue to be a design challenge for future communication systems and advanced multimedia applications. The data compression algorithms can be broadly classified in two categories – lossless and lossy. Usually lossless data compression techniques are applied on text data or scientific data. The discrete wavelet transform (DWT) is being increasingly used for image coding. It is due to the fact that DWT supports superior features like progressive image transmission by quality or by resolution. The DWT is the key component of the JPEG2000 system, [1] and it also has been adopted as the transform coder in MPEG-4 still texture coding. However, the DWT requires much more computation than the discrete cosine transform (DCT) because of filter computation. Recently, lifting scheme widely used for DWT leads a speed-up and a fewer computation compared to the classical convolution-based method. Daubechies and Sweldens first derive the lifting-based discrete wavelet transform to reduce complex operations [2][3]. The lifting-based DWT has several advantages including entire parallel operations, “in-place” computations of the DWT, integer-to-integer transform, symmetric forward and inverse transform, etc. Hence, the lifting scheme is deservedly adopted in the JPEG 2000 image standard. Several efficient DWT architectures are presented by using the lifting-scheme. In general, 2-D DWT is realized by directly executing the 1-D DWT row by row and then column by column [4][5]. However, the huge frame memory is required to store the intermediate coefficients [4][5]. Due to the N2 size of the frame memory, it is usually devised to be the external memory of the DWT chip. Thus, the high external memory bandwidth leads to great power consumption in the 2-D DWT architecture.
S. Jayachandranath, P. Suresh Babu / International Journal of Engineering Research and
Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 6, November- December 2012, pp.1002-1006
1003 | P a g e
The line-based architecture proposed in [6] is an
alternative method to eliminate the frame memory
by performing 1-D DWT in both directories
simultaneously. In order to reduce the external
memory access, the line-based method requires
some internal line buffer to store the intermediate
coefficients. Tseng et al. [7] focus their idea on
optimizing the internal line buffer size for the linebased
2-D DWT architectures. Moreover, several
line-based architectures are proposed based on the
factorized lifting scheme [8]-[15]. Andra et al.[8]
present a lifting-based forward and inverse DWT
architecture with a general 1-D DWT core to
support various DWT filters in JPEG 2000. A
systematic method with systolic array mapping is
proposed to construct several efficient architectures
for 1-D and 2-D lifting-based DWT [9]. Liao et al.
[10] propose two DWT architectures with recursive
and dual scan methods for multi-level and singlelevel
2-D DWT decomposition, respectively.
DWT
DWT analyzes the data at different
frequencies with different time resolutions [1]. Fig.
1 shows the DWT decomposition of the image. The
DWT decomposition involves low-pass „l‟ and highpass
„h‟ filtering of the images in both horizontal
and vertical directions. After each filtering, the
output is down-sampled by two. Further
decomposition is done by applying the above
process to the LL sub-band.
2.2 Lifting Scheme
The lifting scheme has been developed by
Sweldens [4] as an easy tool to construct the second
generation wavelets. The scheme consists of three
simple stages: split, predict (P) and update (U). In
the split stage, the input sequence xj,i is divided into
two disjoint set of samples, even indexed samples
(even samples) xj,2i and odd indexed samples (odd
samples) xj,2i+1. In the predict stage, even samples
are used to predict the odd samples based on the
correlation present in the signal. The differences
between the odd samples and the corresponding
predicted values are calculated and referred to as
detailed or high-pass coefficients, dj-1,i. The update
stage utilizes the key properties of the coarser
signals i.e. they have the same average value of the
signal. In this stage, the coarse or low-pass
coefficient xj-1,i is obtained by updating the even
samples with detailed coefficient. The block
diagram of the lifting based DWT is shown in Fig. 2
ARITHMETIC IN LIFTING DWT
The lifting scheme provides many
advantages, such as fewer arithmetic operations, inplace
implementation and easy management of
boundary extension compared to convolution based
DWT architectures. For simplicity, we use the
popular bi-orthogonal wavelet (5,3) filter, adopted
in JPEG 2000, in order to explain the redundancy in
the arithmetic operation involved in the calculation
of the lifting-based DWT computation. The
calculation of high-pass and low-pass coefficients
for two consecutive values for (5,3) wavelet is
shown below:
[attachment=51591]
Abstract
Image compression is the application of Data compression on digital images. A fundamental shift in the image compression approach came after the Discrete Wavelet Transform (DWT) became popular. To overcome the inefficiencies in the JPEG standard and serve emerging areas of mobile and Internet communications, the new JPEG2000 standard has been developed based on the principles of DWT. In this research, an architecture that performs both forward and inverse lifting-based discrete wavelet transform is proposed. The proposed architecture reduces the hardware requirement by exploiting the redundancy in the arithmetic operation involved in DWT computation. The proposed architecture does not require any extra memory to store intermediate results. Using Verilog HDL, the encoder for the image compression employing DWT was implemented. This architecture has been described in VHDL at the RTL level and simulated successfully using ModelSim simulation environment. In this paper, high-efficient lifting-based architectures for the 5/3 discrete wavelet transform (DWT) are proposed. The proposed parallel and pipelined architecture consists of a horizontal filter (HF) and a vertical filter (VF). The system delays of the proposed architectures are reduced. Filter coefficients of the biorthogonal 5/3 wavelet low-pass filter are quantized before implementation in the high-speed computation hardware. In the proposed architecture, all multiplications are performed using less shifts and additions. The proposed architecture is 100% hardware utilization and ultra low-power. The architecture has regular structure, simple control flow, high throughput and high scalability. Thus, it is very suitable for new generation image compression systems, such as JPEG- 2000. Keywords: 5/3 discrete wavelet transform (DWT), IDWT, horizontal filter (HF), vertical filter (VF), lifting-based architecture, JPEG-2000. 1. INTRODUCTION Data compression is the technique to reduce the redundancies in data representation in order to decrease data storage requirements and hence communication costs. Reducing the storage requirement is equivalent to increasing the capacity of the storage medium and hence communication bandwidth. Thus the development of efficient compression techniques will continue to be a design challenge for future communication systems and advanced multimedia applications. The data compression algorithms can be broadly classified in two categories – lossless and lossy. Usually lossless data compression techniques are applied on text data or scientific data. The discrete wavelet transform (DWT) is being increasingly used for image coding. It is due to the fact that DWT supports superior features like progressive image transmission by quality or by resolution. The DWT is the key component of the JPEG2000 system, [1] and it also has been adopted as the transform coder in MPEG-4 still texture coding. However, the DWT requires much more computation than the discrete cosine transform (DCT) because of filter computation. Recently, lifting scheme widely used for DWT leads a speed-up and a fewer computation compared to the classical convolution-based method. Daubechies and Sweldens first derive the lifting-based discrete wavelet transform to reduce complex operations [2][3]. The lifting-based DWT has several advantages including entire parallel operations, “in-place” computations of the DWT, integer-to-integer transform, symmetric forward and inverse transform, etc. Hence, the lifting scheme is deservedly adopted in the JPEG 2000 image standard. Several efficient DWT architectures are presented by using the lifting-scheme. In general, 2-D DWT is realized by directly executing the 1-D DWT row by row and then column by column [4][5]. However, the huge frame memory is required to store the intermediate coefficients [4][5]. Due to the N2 size of the frame memory, it is usually devised to be the external memory of the DWT chip. Thus, the high external memory bandwidth leads to great power consumption in the 2-D DWT architecture.
S. Jayachandranath, P. Suresh Babu / International Journal of Engineering Research and
Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 6, November- December 2012, pp.1002-1006
1003 | P a g e
The line-based architecture proposed in [6] is an
alternative method to eliminate the frame memory
by performing 1-D DWT in both directories
simultaneously. In order to reduce the external
memory access, the line-based method requires
some internal line buffer to store the intermediate
coefficients. Tseng et al. [7] focus their idea on
optimizing the internal line buffer size for the linebased
2-D DWT architectures. Moreover, several
line-based architectures are proposed based on the
factorized lifting scheme [8]-[15]. Andra et al.[8]
present a lifting-based forward and inverse DWT
architecture with a general 1-D DWT core to
support various DWT filters in JPEG 2000. A
systematic method with systolic array mapping is
proposed to construct several efficient architectures
for 1-D and 2-D lifting-based DWT [9]. Liao et al.
[10] propose two DWT architectures with recursive
and dual scan methods for multi-level and singlelevel
2-D DWT decomposition, respectively.
DWT
DWT analyzes the data at different
frequencies with different time resolutions [1]. Fig.
1 shows the DWT decomposition of the image. The
DWT decomposition involves low-pass „l‟ and highpass
„h‟ filtering of the images in both horizontal
and vertical directions. After each filtering, the
output is down-sampled by two. Further
decomposition is done by applying the above
process to the LL sub-band.
2.2 Lifting Scheme
The lifting scheme has been developed by
Sweldens [4] as an easy tool to construct the second
generation wavelets. The scheme consists of three
simple stages: split, predict (P) and update (U). In
the split stage, the input sequence xj,i is divided into
two disjoint set of samples, even indexed samples
(even samples) xj,2i and odd indexed samples (odd
samples) xj,2i+1. In the predict stage, even samples
are used to predict the odd samples based on the
correlation present in the signal. The differences
between the odd samples and the corresponding
predicted values are calculated and referred to as
detailed or high-pass coefficients, dj-1,i. The update
stage utilizes the key properties of the coarser
signals i.e. they have the same average value of the
signal. In this stage, the coarse or low-pass
coefficient xj-1,i is obtained by updating the even
samples with detailed coefficient. The block
diagram of the lifting based DWT is shown in Fig. 2
ARITHMETIC IN LIFTING DWT
The lifting scheme provides many
advantages, such as fewer arithmetic operations, inplace
implementation and easy management of
boundary extension compared to convolution based
DWT architectures. For simplicity, we use the
popular bi-orthogonal wavelet (5,3) filter, adopted
in JPEG 2000, in order to explain the redundancy in
the arithmetic operation involved in the calculation
of the lifting-based DWT computation. The
calculation of high-pass and low-pass coefficients
for two consecutive values for (5,3) wavelet is
shown below: