A heterogeneous image fusion algorithm based on LLC coding

Abstract

Most image fusion algorithms are not good at batch processing. To address this, we propose a LLC coding based image fusion algorithm, by which multiple infrared and visible light images can be fused and identified. The images were encoded and several image features were extracted by those codes. It was judged whether the images could be merged by the coincidence of the non-zero coding counterpart obtained from comparing the LLC coding of two heterogeneous images. The max-pooling criterion was employed to fuse the features extracted from images by maximizing the complementary information and minimizing the redundant information. Consequently the SVM classifier was used to classify and identify the target. The simulated results show the accuracy of our proposed method.