The goal of image fusion is to create new images that are more suitable for the purposes of human visual perception, object detection and target recognition. For Automatic Target Recognition (ATR), we can use multi-sensor data including visible and infrared images to increase the recognition rate. In this paper, we propose a new multiresolution data fusion scheme based on Daubechies Wavelet Basis (DWB) and pixel-level weights including thermal weights and visual weights. We use multiresolution decompositions to represent the input images at different scales, present a multiresolution/multimodal segmentation to partition the image domain at these scales. The crucial idea is to use this segmentation to guide the fusion process. Physical thermal weights and perceptive visual weights are used as segmentation multimodals. Daubechies Wavelet (at different levels) is choosen as the Wavelet Basis. Experimental results confirm that the proposed algorithm is the best image sharpening method and can best maintain the spectral information of the original infrared image. Also, the proposed technique performs better than the other ones in the literature, more robust and effective, from both subjective visual effects and objective statistical analysis results.
Forward Looking Infra-Red (FLIR) image segmentation is crucial for Automatic Target Recognition (ATR). This paper presents a thresholding method for image segmentation by performing fuzzy partition on a two-dimensional (2-D) histogram based on maximum entropy principle. We combine the original image with its smooth image to form a binary set, called a "generalized image", and the histogram of the generalized image is a 2-D histogram. In order to adequately utilize the intrinsic information of the FLIR image, we adopt a newly defined fuzzy partition of two fuzzy sets, dark and bright, basing on 2-D histogram. Also we define the corresponding 2-D membership function, which represents the membership of darkness and brightness for each element in the binary set, respectively. The entropy is used as a measure of fuzziness. Based on the Shannon function, we define a 2-D fuzzy entropy. The total fuzzy entropy is the sum of the entropy of each block. Therefore, the fuzzy region can be determined by maximizing the total fuzzy entropy. A genetic algorithm is employed to find the optimal combination of all the fuzzy parameters. Experiment results show that the proposed method gives good performance.