INTERNATIONAL JOURNAL FOR TRENDS IN ...ijtet.com/wp-content/plugins/ijtet/file/upload/docx/5968...The proposed hyperspectral image compression is based on lifting wavelet transform

INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY

VOLUME 14 ISSUE 1 – JUNE 2016 - ISSN: 2349 – 9303

27

Implementation Based Compression of Hyper spectral

Images Using Lifting Transform

Mahendran. M 1

Department of ECE,

National Engineering College,

Kovilpatti, India

[email protected]

Jayavathi. S.D2

Department of ECE,

National Engineering College,

Kovilpatti, India

[email protected]

Abstract– Hyperspectral images are a satellite image. It contains a lot of information and set of bands. The HSI compression is an

important issue in remote sensing application. The proposed hyperspectral image compression is based on lifting wavelet transform

(LWT).In this project, the image file is convert into a header file, it contain a pixels value of the image. And then it is followed by 2D

LWT is applied to the wavelet coefficient for hyperspectral image compression. The proposed method implemented on Xilinx 10.1,

Spartan3-EDK kit and tested the hyperspectral image. Experiment on data from the urban data is analyzed. Hyperspectral image

compression is a lossy image compression technique and it provides good computational speed. Finally the compressed output image is

display the visual basic tool.

Index Terms— Compression, hyperspectral images, 2D lifting wavelet transform.

—————————— ——————————

1. INTRODUCTION

yperspectral images are widely used in a variety of fields,

such as target detection, material identification, ground

mapping and agriculture. The advancement of sensor

technology produces remotely sensed data that have a large

number of spectral bands. There is an increasing need for

efficient compression techniques for these hyperspectral images.

The compression of hyperspectral images can be implemented

by detecting the spatial and spectral redundancies. The

compression methods can be classified into two types: lossless

compression and lossy compression. A lossless technique that

decompresses data back to its original from without any loss.

Redundant data is removed from compression and added to

decompression. Lossless compression methods are

recommended in hyperspectral images due to the huge quality of

data and the data loss must be small. Most the lossy compression

methods resort to transform based approaches. In particular

transformed based methods, principal component analysis has

commonly used, often followed by 2-D transform such as the

DWT or DCT. Several methods expand known two dimensional

transform based methods into 3D applications, including SPIHT

(set partitioning in hierarchical trees), SPECK (set partitioning

embedded block). Most lossy compression methods are

developed to minimize mean squared errors between original

and reconstruct the pixels.

In this paper, we propose the image file is converting a

header file, and it is followed by 2D LWT is applied. Finally

implement on FPGA spartan3 and different image result is

analyzed.

2. IMAGE COMPRESSION METHODS

2.1Conversion of image to Header File

Matrix can be used to represent the image, the data type

of matrix is unit 8, and each element of matrix corresponds to a

pixel image, these pixel values are in the interval of [0, 255].

Figure 1 is the example for image to pixels values.

Figure 1 : Block Diagram of Proposed Method

2.2 Proposed Compression Technique

Color space conversion is a converts RGB image to

Grayscale image by eliminating the hue and saturation .

H

INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY

VOLUME 14 ISSUE 1 – JUNE 2016 - ISSN: 2349 – 9303

28

Figure 2 : Block Diagram of Proposed Method

C. Wavelet Transform:

The wavelet transform decomposes a signal into a set of

basics function. The wavelet lifting scheme is a method for

decomposing wavelet transform into a set of stages. This method

based on haar wavelet.

The lifting scheme is an alternative method of

computing the wavelet coefficients.

Advantages of the lifting scheme:

Requires less computation time and less memory.

Linear, nonlinear, and adaptive wavelet transform is

feasible, and the resulting transform is invertible and

reversible.

A spatial domain construction of bi-orthogonal wavelets

consists of the 4 operations:

Split: Splitting the signal into two parts is called lazy wavelets,

because we have not performed any mathematical operations. It

divides input data into odd and even function.

𝑆𝐾(0)

= 𝑋2𝑖(0)

; 𝑑𝑘(0)

= 𝑋2𝑖+1(0)

(2)

Predict: The even and odd samples are interleaved. If the signal

is having locally correlated structure, then even and odd samples

are highly correlated. It is very easy to predict odd samples from

even samples.

𝑑𝑘(𝑟)

= 𝑑𝑘(𝑟−1)

− 𝑃𝑗𝑟 𝑆𝑘+𝑗(𝑟−1)

(3)

Update: The coarser signal must have same average value that

of original signal. To do this, we require lifting the λ -1, k with

help ofwavelet coefficients γ-1, k. After lifting process, mean

value of original signal and transformed signal remains same.

We require constructing update operator U for this lifting

process.

𝑟 = 𝑆𝑘 𝑟 − 1 + 𝑈𝑗𝑟 𝑑𝑘+𝑗(𝑟)

(4)

Figure 3: Lifting Scheme

3. HARDWARE DESCRIPTION

Figure 4: TYRO PLUS SPARTAN3 (EDK) Board

The Spartan-3 EDK Board provides a powerful, self-

contained development platform for designs targeting the new

Spartan-3 FPGA from Xilinx. It features a 200k gate Spartan-3,

on-board I/O devise, and 1MB fast asynchronous SRAM,

making it the perfect platform to experiment with any new

design, from a simple logic circuit to an embedded processor

core. The board also contains a platform Flash JTAG-

programmable ROM, so designs can easily be made non-

volatile.

The Spartan-3Starter Board is fully compatible with all

versions of the Xilinx ISE tools, including the free web pack.

The board ships with a power supply, and programming cable.

INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY

VOLUME 14 ISSUE 1 – JUNE 2016 - ISSN: 2349 – 9303

29

Figure5: Block Diagram of Xilinx Spartan-3 EDK

Static random-access memory (SRAM) is a type of

semiconductor memory that uses bitable latching circuitry to store

each bit. The SRAM array forms either a single 256Kx32 SRAM

memory or two independent 256Kx16 arrays. Both SRAM

devices share common write-enable, output-enable and address

signals. However, each device has a separate chip select enable

control and individual byte-enable controls to select the high or

low byte in the 16-bit data word.

The 256Kx32 configuration is ideally suited to hold

Micro Blaze instructions. However, it alternately provides high-

density data storage for a variety of applications, such as digital

signal processing (DSP), large data FIFOs, and graphics buffers.

4. EXPERIMENTAL RESULTS

The hyperspectral images compression steps of urban

image use the compression method based on 2D LWT encoding

decoding and are shown in Figure 6(a), the input hyperspectral

image is formed by combining the bands 85, 30, 110 for the

RGB colors respectively. Figure6(b), shows the grayscale image

(255x255) of the urban data, and Figure7 shows the conversion

of input image to header file, that header file (image.h) contains

pixels value of the input image.

(a) (b)

Figure 6: Hyperspectral Images: (a) Input urban image

(b) Grayscale Image

Figure 7: Conversion of input image to header file

INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY

VOLUME 14 ISSUE 1 – JUNE 2016 - ISSN: 2349 – 9303

30

Figure 8: Proposed Spartan3 EDK Kit

Figure 8 shows the proposed hardware spartan3 EDK kit.

Next power on the board and the system can be initialized by

update the hardware bit stream. This updated bit stream can then

be downloaded to the FPGA. Under device configuration, choose

download bit stream. The bit stream is initialized with the

executable. The arrows are specifying the program download

successfully.

Figure 9: Snapshot for Xilinx Platform Studio

Figure 10: Snapshot for Launch XMD

Figure 9 shows the snapshot for Xilinx platform studio; Figure

10 shows the launch XMD of the debug steps, first type the CD

filename and ‗dowexecutable.elf‘ in command window. It is

download the program and execute the device configuration,

finally run the program.

Figure 11: lifting wavelet transforms (2D)

Figure 11 shows lifting wavelet transform of the output

image for cameraman. Figure 12 shows the LWT first

decomposition of the output image for hyperspectral image.

Figure 12: 1D Lifting wavelet transform for an hyperspectral

image

The performance of the compression techniques is

analyzed and compared in terms of computational time and

compression ratio.

CR = 𝒏𝒐.𝒐𝒇 𝒃𝒊𝒕𝒔 𝒊𝒏 𝒐𝒓𝒈𝒊𝒏𝒂𝒍 𝒊𝒎𝒂𝒈𝒆

𝒏𝒐 𝒐𝒇 𝒃𝒊𝒕𝒔 𝒊𝒏 𝒄𝒐𝒎𝒑𝒓𝒆𝒔𝒔𝒆𝒅 𝒊𝒎𝒂𝒈𝒆 (5)

INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY

VOLUME 14 ISSUE 1 – JUNE 2016 - ISSN: 2349 – 9303

31

5. CONCLUSION An improved method is proposed for the implementation

based compression of hyperspectral images using lifting wavelet

transform. First we applied PCA to the input image for extracting

the features and then it is followed by lifting wavelet transform.

First the image file is convert into a header file, it contain a pixels

value of the image. And then it is followed by 1D LWT is applied

to the wavelet coefficient for hyperspectral image compression. It is

lossy compression technique, it provides good computational

speed. Finally the compressed output image is obtained by taking

inverse lifting wavelet transform to the image. The performance

of the output image is analyzed. In future, the above algorithm will

be implementation based two dimensional wavelet transform on

different size of images.

REFERENCES

[1]. S.Kala, Dr.S.Vasuki, ―FPGA Based Hyperspectral image

Compression Using DWT and DCT‖, Australian journal of

basic and applied sciences, May 2014.

[2]. Chulhee Lee, SungwookYoun, TaeukJeong, Eunjae Lee, and Joan

Serra-Sagristà, ―Hybrid Compression of Hyper spectral Images

Based on PCA with Pre-Encoding Discriminant Information‖,

IEEE geosciences and remote sensing letters, vol. 12, no.

7, July 2015.

[3]. Merav Huber-Lerner, OferHadar, Senior Member, IEEE, Stanley

R. Rotman, Member, IEEE, and Revital Huber-Shalem,

―Compression of Hyper spectral Images Containing a Sub pixel

Target‖, IEEE journal of selected topics in applied earth

observations and remote sensing, vol. 7, no. 6, June

2014 .

[4]. A.AliceBlessie, J. Nalini and S.C.Ramesh, ―Image Compression

Using Wavelet Transform Based on the Lifting Scheme and its

Implementation‖, IJCSI International Journal of Computer

Science Issues, Vol. 8, Issue 3, No. 1, May 2011.

[5]. William A. Pearlman, Asad Islam, Nithin Nagaraj, and Amir

Said, ―Efficient, Low-Complexity Image Coding with a Set-

Partitioning Embedded Block Coder‖, Bellingham, WA, USA:

SPIE, Nov. 2013.

[6]. S.Narasimhulu, Dr.T.Ramashri, ―Gray-Scale Image Compression

Using DWT-SPIHT Algorithm‖, IJERA, August 2012.

[7]. Lena Chang, Yang-Lang Chang, Z. S. Tang, and Bormin

Huang,―Group and Region Based Parallel Compression Method

Using Signal Subspace Projection and Band Clustering for Hyper

spectral Imagery‖, IEEE journal of selected topics in applied

earth observations and remote sensing, vol. 4, no. 3,

September 2011.

[8]. Qian Du, and James E. Fowler, ―Hyperspectral Image

Compression Using JPEG2000 and Principal Component

Analysis‖, IEEE geosciences and remote sensing letters,

vol. 4, no. 2, April 2007.

[9]. X. Tang and W. A. Pearlman, ―Three-dimensional wavelet-based

compression of images in Hyperspectral Data Compression‖, G.

Motta, F. Rizzo, and J. A. Storer, Eds. Norwell, MA, USA:

Kluwer, April 2006.

[10].William A. Pearlman, Asad Islam, Nithin Nagaraj, and Amir

Said, ―Efficient, Low- Complexity Image Coding With a Set-

Partitioning Embedded Block Coder IEEE transactions on

circuits and systems for video technology, vol. 14, no.11,

November 2004.

[11].P.L.Dragotti, G. Poggi, and A. R. P. Ragozini, ―Compression of

multispectral images by three-dimensional SPIHT algorithm,‖

IEEE Trans.Geosci. Remote Sens., vol. 38, no. 1, pp. 416–428,

Jan. 2000.