Pixel-level Visual and Thermal Images Fusion Using Maximum and Minimum Value Selection Strategy

In this paper, an effective pixel-level fusion algorithm is presented for visible and thermal images obtained from different sensors. The fused image may contain more essential information for human as well as machine intuition. Because of the said merits, pixel-level image fusion plays an essential role in military, remote sensing, and night vision applications. Initially, a center sliding window is used in the infrared image. The initial feature map is obtained by utilizing the maximum selection strategy between the pixel-intensity of visible image and block-based average of center sliding window. This initial feature map is not able to preserve the useful information, which is applicable for human visible intuition. So in the next stage, the source images are compared by using the minimum selection strategy to obtain the final feature map. Eventually, the fused image is achieved by utilizing the weighted-average technique among the feature maps. To estimate the efficiency of the proposed algorithm quantitatively, and qualitatively tests were carried out on a publicly available “TNO-database”. The experimental results of the proposed algorithm attained higher accuracy as compared to seven existing techniques in visual quality and objective assessment.