A multi-algorithm block fusion method based on set-valued mapping for dual-modal infrared images

It is always a goal to make use of difference-features between infrared intensity and polarization (degree of polarization) images to drive selection of targeted algorithms rather than using fixed algorithm, and improve the fusion efficiency under the premise of ensuring the high-quality fusion performance in the field of infrared images fusion. In this paper, with the source of infrared intensity and polarization images, a multi-algorithm block fusion method is proposed based on set-valued mapping relationship between the difference-features and the optimal fusion algorithms. Firstly, by comparing the basic imaging process, the respective characteristics of dual-modal infrared images are identified, and then the salient difference-features are selected according to their image characteristics and complementary information. Further, the optimal fusion algorithm for each difference-feature is developed by using the proposed fusion validity formula. Based on the above, we establish the probability distribution between the difference-features set and the optimal fusion algorithms set, and further determine the set-valued mapping relationship between them. Second, the source images are divided into blocks and the main difference-feature of each image block is extracted, then the optimal fusion algorithm is selected according to the established mapping relationship for block fusion. Finally, the fused image blocks are stitched into a complete fusion image by the corresponding splicing processing. Through experimental comparison, the proposed method not only improves the fusion quality, but also has satisfactory work efficiency. In addition, the proposed method can fully exploit the ability of difference-features to drive selecting optimal fusion algorithm