Introduction Remote Sensing Image Fusion Scheme using Directional Vector in NSCT Domain

TELKOMNIKA, Vol.14, No.2, June 2016, pp. 598~606 ISSN: 1693-6930, accredited A by DIKTI, Decree No: 58DIKTIKep2013 DOI: 10.12928TELKOMNIKA.v14i1.2747  598 Received December 27, 2015; Revised March 18, 2016; Accepted April 4, 2016 Remote Sensing Image Fusion Scheme using Directional Vector in NSCT Domain Baohui Tian 1 , Lan Lan 1 , Hailiang Shi 2 , Yunxia Pei 2 1 Department of Communication Information Engineering, Henan Vocational Technical College of Communications, Zhengzhou 450005, Henan, China 2 College of Mathematics Information Science, Zhengzhou University of Light Industry, Zhengzhou 450002, Henan, China Corresponding author, e-mail: 18341615qq.com Abstract A novel remote sensing image fusion scheme is presented for panchromatic and multispectral images, which is based on NonSubsampled Contourlet Transform NSCT and Principal Component Analysis PCA. The fusion principles of the different subband coefficients obtained by the NSCT decomposition are discussed in detail. A PCA-based weighted average principle is presented for the lowpass subbands, and a selection principle based on the variance of the directional vector is presented for the bandpass directional subbands, in which the directional vector is assembled by the NSCT coefficients of the different directional subbands but the same coordinate. The proposed scheme is tested on two sets of remote sensing images and compared with some traditional multiscale transform-based image fusion methods, such as discrete wavelet transform, stationary wavelet transform, dual-tree complex wavelet transform, contourlet transform. Experimental results demonstrate that the proposed scheme provides superior fused image in terms of several relevant quantitative fusion evaluation indexes. Keywords: Image Fusion, Remote Sensing, Nonsubsampled Contourlet Transform, Principal Eigenvector, Directional Vector Copyright © 2016 Universitas Ahmad Dahlan. All rights reserved.

1. Introduction

The field of remote sensing is a continuously growing market with applications like vegetation mapping and observation of the environment. The increase in applications is due to the availability of high quality images for a reasonable price and improved computation power. However, as a result of the demand from higher classification accuracy and the need in enhanced positioning precision there is always a need to improve the spectral and spatial resolution of remote sense images. These requirements can be either fulfilled by building new satellites with a superior resolution power, or by the utilization of image fusion techniques. The main advantages of the second alternative is the significantly lower expense [1]. In remote sensing systems, panchromatic PAN images of high spatial resolution can provide detailed geometric information, such as shapes, features and structures of objects of the earth’s surface, and multispectral MS image with relatively lower spatial resolution are used to obtain spectral information necessary for environmental applications. The different objects within images of high spectral resolution and high spatial resolution are easily identified [2]. The goal of image fusion is to obtain a high spatial resolution multispectral image which combines the spectral characteristic of the low spatial resolution data with the spatial information of the panchromatic images. Unlike other application, e.g. image fusion in military missions or computer-aided quality control, the main constraint in remote sensing is to preserve the spectral information for tasks [1, 2]. In the last decades, many methods have been proposed for fusing panchromatic and multispectral images. The well known methods are those based on the intensity-hue-saturation transform IHS, principal component analysis PCA, discrete wavelet transform DWT and so on [3]. Unfortunately, these methods have their limitations. The main drawback of the methods based on IHS or PCA is the high distortion of the original spectral information [4]. The two dimensional 2-D DWT is good at isolating the discontinuities at edge points, but cannot effectively represent the ‘line’ or the ‘curve’ discontinuities properly [5]. In addition, DWT can TELKOMNIKA ISSN: 1693-6930  Remote Sensing Image Fusion Scheme using Directional Vector in NSCT… Baohui Tian 599