IMAGE FUSION AND IMAGE QUALITY ASSESSMENT OF FUSED IMAGES

It is of great value to fuse a high-resolution panchromatic image and low-resolution multi-spectral images for object recognition. In the paper, tow frames of remotely sensed imagery, including ZY03 and SPOT05, are selected as the source data. Four fusion methods, including Brovey, PCA, Pansharp, and SFIM, are used to fuse the images of multispectral bands and panchromatic band. Three quantitative indicators were calculated and analyzed, that is, gradient, correlation coefficient and deviation. According to comprehensive evaluation and comparison, the best effect is SFIM transformation, combined with fusion image through four transformation methods. * Corresponding author. Han Zhen,Tel:13521798206,E-mail:hanzhen_hz@126.com.


INTRODUCTION
In recent years, the launch of high-resolution satellites has opened a new era for remote sensing.With these remote sensors, images of various spatial and spectral characteristic can be obtained.For example, from the ZY03 sensor, 2.1 m resolution panchromatic and 5.6 m resolution multispectral and 3.5 m resolution front-facing, rear-facing images are available.With the development of remote technique and the launch of new satellites, remote image have great promoted at spatial resolution and spectral resolution (Wang, 2003).Fusion of two images from different sources can provide information that cannot be obtained when the images are processed individually.Image fusion can improve the accuracy when we extract useful information from image (Jia, 2000).To data, various fusion algorithms have been developed ( Chavez, 1991, Siddiqui, 2003, Hallada, 1983, Vrabel, 2002).As Zhang combined DWT and IHS algorithms and apply to image fusion of the IKONOS and Quick Bird (Zhang, 2005).Yang combined NSCT and HIS and apply to image fusion of the SPOT and TM (Yang, 2008).In the paper, we tested Brovey ，PCA，Pansharp and SFIM .And we assessed the quality of fusion image with indices of gradient, correlation coefficient and deviation.

Image fusion methods
In this section, we analyze image fusion method that can be used for high-resolution satellite image fusion, such as those for fusion of panchromatic and multi-spectral images.Four categories of image fusion methods are addressed: Brovey, PCA, Pansharp, and SFIM .

1) Brovey Algorithm
The Brovey Algorithm is a ratio method that normalizes multispectral bands used for a RGB display (GILLESPIE, 1987).
The result is then multiplied by a high-resolution band to add high spatial frequency information.The Brovey Algorithm can only allow a limited number of bands to be fused.The Brovey Algorithm can be defined as follows: gray-value relationship between the original multispectral image, panchromatic image, and fused image.This method also statistics all bands of the input and fusion results in order to eliminate the dependence of the data set (Tan,2008).

4) SFIM Algorithm
The SFIM is a ratio method that the high-resolution image is divided by a simulated low-resolution image and the result is then multiplied by the low-resolution image.The algorithm was defined by LIU (2000a) as below:

Quality indicators for assessing image fusion
The purpose of image fusion is to enhance the spatial and spectral resolution from several low-resolution images.Several indices have been proposed for assessing image quality, which can also be applied to assessing the quality of a fusing image.In this section, we select three indices to assess the quality of fusion image.Those includes: gradient, correlation coefficient and deviation.1) Gradient: The gray-scale variation of the image in a direction.Tiny detail and texture changes can be reflected with this index.2) Deviation: where F and R are two images,   j i R , and   j i F , the elements of the image R and the image F, respectively.R and F stand for their mean values.If C (F, R) is closer to 1, which means that the two images are more similar and the better the fusion image fidelity.

Result
In the paper, four fusion methods, including Brovey, PCA, Pansharp, and SFIM , are used to fuse the images of multispectral bands and panchromatic band.Three quantitative indicators were calculated and analyzed, that is, gradient, correlation coefficient and deviation.

Quantitative analysis
Considering the drawbacks of the subjective quality assessment method, much effort has been devoted to develop objective image quality assessment methods.A good fusion approach retain the maximum spatial and spectral information from the original images and should not damage the internal relationship among the original bands.Based on these three criteria, we select gradient, correlation coefficient and deviation to assess the fusion image.13.102892 ≈ Pansharp: 13.163267), the value of deviation obtained by SFIM transformation method is minimum, and the value of correlation coefficient obtained by SFIM transformation method is the largest.From these quantitative results, we can conclude that the fusion result which is obtained by SIFM transformation method is the best.Form Table 2,we observe that the value of gradient obtained by SFIM transformation method is the largest, the value of deviation obtained by SFIM transformation method is minimum, and the value of correlation coefficient obtained by SFIM transformation method is the largest.From these quantitative results, we can conclude that the fusion result which is obtained by SIFM transformation method is the best.

CONCLUS IONS
Selection of proper fusion technique depends on the specific remote image.Four fusion methods, including Brovey, PCA, Pansharp, and SFIM , are used to fuse the images of multispectral bands and panchromatic band.Three quantitative indicators were calculated and analyzed, that is, gradient, correlation coefficient and deviation.Finally, form the above analysis and comparison, we can conclude that SFIM algorithm can preserve the spectral characteristics of the source multispectral image as well as the high spatial resolution characteristics of the source panchromatic image and suited for fusion ZY03 and SPOT05 images.
one of the bands, that is red band, green band or blue band.value of red band, green band and blue band.h B is the high-resolution image.2) PCA Algorithm PCA transform is created based on the statistical characteristics dimensional linear transformation, the mathematical transformation is called K-L(Sun, 2002).The PCA transform converts intercorrelated M S bands into a new set of uncorrelated components.The first component also resembles a PAN image.It is, therefore, replaced by a high-resolution PAN for the fusion.The PAN image is fused into the low-resolution M S bands by performing a reverse PCA transform.3)Pansharp AlgorithmThe PANSHARP algorithm is based on the least number of squares to an approximate International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-7/W1, 3rdISPRS IWIDF 2013, 20 -22 August 2013, Antu, Jilin Province, PR China in the i th band of a low-resolution input image, co-registered to a high-resolution input image high B , and mean B is a corresponding pixel in a simulated lowresolution image derived from high B using an average filter for a neighborhood equivalent to the resolution of high B .
is the deviation index of the k th band, k j i DN , denotes the original image gray level of a pixel with coordinate   j i, .
the fusion image gray level of a pixel with coordinate   j i, .Deviation index used to compare the degree of deviation of fusion images and low-resolution multispectral images.3) Correlation: Figure.1 is the original image of ZY03, Figure.2 is the fusion image of ZY03, Figure.3 is the original image of SPOT05, Figure.4 is the fusion image of SPOT05the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-7/W1, 3rd ISPRS IWIDF 2013, 20 -22 August 2013, Antu, Jilin Province, PR China

Table 2
Assessment results of SPOT05 fused imageForm Table1,we observe that the value of gradient obtained by SFIM transformation method is almost the largest(SFIM :