The information content of a single image is mainly limited by the spatial and spectral resolution of the imaging system. Current imaging systems somehow offer a trade-off between high spatial and high spectral resolution no single system offers both of these characteristics. For example, a high spatial resolution system such as IKONOS, can supply 1m spatial resolution images but the spectral information is limited to a single panchromatic (0.45μm – 0.90μm) band. The sensor also supplies 4 multi-spectral (blue, green, red and near infra-red) bands, i.e. with high spectral resolution, but with low spatial resolution of 4 meters. However, the image processing methods can overcome this limitation. In order to obtain the both characteristics in a single image, that is high spatial and high spectral resolution, a technique image fusion can be employed. Work in this field has marked a satisfactory outcome in last few years but it has not defined or estimated the quality of information improvement quantitatively. This paper presents the assessment of image fusion by measuring the quantity of enhanced information in fused images. Two measuring methods – Entropy and Image Noise Index were employed. Entropy can measure the information content of images but it has a limitation. It cannot distinguis h between information and noise. A solution to this limitation is discussed and a new method is proposed – the Image Noise Index (INI) using entropy. This method was applied on three commonly used image fusion techniques intensity-hue-saturation (IHS), principal component analysis (PCA) and high pass filter (HPF). The INI showed definite results distinguishing between noise and information. It also compares the fusion techniques and indicates which technique gives better results. 1. INTRODUTION Image fusion is a useful technique for merging similar-sensor and multi-sensor images to enhance the image information. The purpose of multi-sensor fusion is to synthesize different pieces of image data coming from different sensors into a single data set. Other terms such as the enhancement of the spatial resolution of multispectral image or to sharpen the multi-spectral image or to merge different information from different sensors are used to describe multi-sensor fusion. Multi-sensor fusion is more convenient and economical than designing an advanced sensor with both resolution characteristics. Previous works (Carper et al., 1990; Chavez et al., 1991 and Munechika et al., 1993) have recognized that multisensor fusion can achieve the child image with more information than either parent image. Their assessment was based on visual and graphical inspection and they have not defined or estimated the quality and degree of information improvement quantitatively. Recently, Pohl and Van Genderen (1998) have reviewed this topic in detail. They emphasized future works on methods to estimate and assess the quality of fused imagery. Entropy is a measure of information and its concept has been employed in many scientific fields. It has been applied in image processing methods as a measure of information but has not been used to assess the effects of information change in fused images. The reason is that entropy sees information as a frequency of change in the digital numbers in images. It cannot distinguish between information belonging to the scene and noise. A new method using entropy is developed in this study Image Noise Index (INI) to assess the effects of information change in fused images. Initially, this paper describes entropy as a quantitative measure of information content of an image and INI to assess the information change. SPOT panchromatic and multi-spectral SPOT XS image data are then fused by the IHS (Intensity-Hue-Saturation), PCA (Principal Component Analysis) and HPF (High Pass Filter) fusion approaches and used as examples to explain the use of entropy and the INI. Finally, the benefits of the INI and its limitations in fusion assessment are discussed.
[1]
Mohammad F. Daemi,et al.
Recognition information
,
1996,
Proceedings of 13th International Conference on Pattern Recognition.
[2]
S. Sides,et al.
Comparison of three different methods to merge multiresolution and multispectral data: Landsat TM and SPOT panchromatic
,
1991
.
[3]
J. G. Liu,et al.
Smoothing Filter-based Intensity Modulation : a spectral preserve image fusion technique for improving spatial details
,
2001
.
[4]
J. Schott,et al.
Resolution enhancement of multispectral image data to improve classification accuracy
,
1993
.
[5]
J. G. Liu,et al.
Smoothing Filter-based Intensity Modulation: A spectral preserve image fusion technique for improving spatial details
,
2000
.
[6]
Christine Pohl,et al.
Multisensor image fusion in remote sensing: concepts, methods and applications
,
1998
.
[7]
Sang Joon Kim,et al.
A Mathematical Theory of Communication
,
2006
.
[8]
J. C. Price,et al.
Comparison of the Information Content of Data from the LANDSAT-4 Thematic Mapper And the Multispectral Scanner
,
1984,
IEEE Transactions on Geoscience and Remote Sensing.
[9]
W. J. Carper,et al.
The use of intensity-hue-saturation transformations for merging SPOT panchromatic and multispectral image data
,
1990
.