A Local Feature Based Fusion Algorithm for Fire Detection Image

Image fusion is an very important process for fire detection, it generates a single combined image that contains a more accurate description of the fire scene than multiple images from different sources. In this paper, a local feature based fusion algorithm for fire detection image is proposed. Using the non-subsampled contourlet transform (NSCT), each of the visual and infrared fire detection images is decomposed into a low frequency and a set of high frequency subbands. Then the fused coefficients are generated by applying different local feature measure to the low frequency and high frequency subbands, respectively. And the fused image is obtained by taking inverse NSCT. Experimental results indicate that the proposed method outperforms other methods in both of fire target enhancement and background detail preservation.