Chinese Journal of Computers   Full Text
Title	Synthesis Performance Evaluation of Multi-Sensor Image Fusion
Authors	HE Gui-Qing CHEN Shi-Hao TIAN Yun HAO Chong-Yang
Address	(Institute of Electronic Information Engineering of Northwestern Polytechnical University, Xi’an 710072) (Sino-German Institute of Information Technology of Northwestern Polytechnical University, Xi’an 710072)
Year	2008
Issue	No.3(486—492)
Abstract &   Background	Abstract To resolve the evaluation issue of image fusion, this paper proposes a novel multi-hierarchical synthesis approach using the Grey Relational Analysis mechanism which has the merits of small-sized sample and unitary comparing. The proposed approach combines the apriori knowledge and quantization evaluation. This paper first gives a six-step procedure for performance evaluation of a single level of hierarchical grey relational analysis and then, based on the single level analysis, gives a four-step procedure for performance evaluation of hierarchical grey relational analysis. Therefore it gets the synthetic evaluation results, which is more quantitative and comprehensive than the conventional objective measures such as correlation coefficient and average gradient. The novel approach gives not only overall performance but also required special performance evaluation. Extensive experimental analysis shows that the proposed approach performs better on quantization, precision, objectivity, reliability and real-time. The advantages above make it apply properly to fusion system with feedback capability and it enriches and perfects the image fusion system. keywords image fusion；performance evaluation；synthesis method；hierarchical；grey relational analysis background As an important branch of multi-sensor information fusion, image fusion has found wide applications in many areas such as remote sensing, medicine, weather forecast and military object identification etc. During the past years, many algorithms such as IHS transform, PCA, HPF, wavelet transform and their modifications have been proposed to solve the fusion issue of multi-spectra, multi-phase, multi-resolution and multi-sensor problems. However, research on the all-sided and objective performance evaluation of image fusion is relatively less, though it can make a comparison among these different fusion algorithms, and thus to provide instructions on both the modification of presents algorithms and the proposal of new algorithms. Current basic method for evaluation system of image fusion usually includes subjective and objective evaluation. Subjective evaluation is mainly through visual judging, that is to say, one looks at the image and gives out the results such as: Is the image much more clearer? Is there any distortion in image spectrum? Objective evaluation is executed mainly through the combination of statistics analysis such as correlation coefficient, variance and entropy etc. By contrast, the developing method for evaluation system, synthesis evaluation, integrates the objective criteria with non-linear theories or intelligent computing algorithms. In this paper, a novel multi-hierarchical synthesis evaluation approach is proposed, using the Grey Relational Analysis which has the merits of small-size sample and unitary comparing mechanism. The proposed approach combines the apriori knowledge and quantization evaluation, gives not only overall performance but also required special performance evaluation. Extensive experimental analysis shows that the proposed approach performs better on quantization, precision, objectivity, reliability and real-time. This work is supported by a grant for the Ph.D. Programs Foundation of Ministry of Education of China under grant No.20040699015 and Youth for NWPU teachers Scientific and Technological Innovation Foundation under grant No.5210102-0800-M016206.