Region-based multisensor image fusion method

1 .INTRODUCTIONI mage fusion can be defined as the process by whichseveral i mages , or some of their feature , are com-binedtogether to formdifferent modalities or instru-ments. It is of great i mportance in many applica-tions ,such as object detection, ATR(automatic tar-get recognition) , remote sensing, computer visionand robotics .There are various techniques for i mage fusion.The selection of the appropriate one depends stronglyon the type of application. Asi mple approach for fu-si…

Region-based multisensor image fusion method

Image fusion should consider the priori knowledge of the source images to be fused, such as the characteristics of the images and the goal of image fusion, that is to say, the knowledge about the input data and the application plays a crucial role. This paper is concerned on multiresolution (MR) image fusion. Considering the characteristics of the multisensor (SAR and FLIR etc) and the goal of fusion, which is to achieve one image in possession of the contours feature and the target region feature. It seems more meaningful to combine features rather than pixels. A multisensor image fusion scheme based on K means cluster and steerable pyramid is presented. K means cluster is used to segment out objects in FLIR images. The steerable pyramid is a multiresolution analysis method, which has a good property to extract contours information at different scales. Comparisons are made with the relevant existing techniques in the literature. The paper concludes with some examples to illustrate the efficiency of the proposed scheme.