A Multi-pose Image Fusion Research Based on Structured Block and Edge Superposition

摘要

Aiming at the problem of multi-pose image fusion in motion scenes, a fusion method based on structured sub-block decomposition and edge superposition is designed. For the problems of preserving information by structured sub-block decomposition, ghost removal by motion consistency detection, and visual fidelity enhancement by edge superposition, an image fusion model based on structured sub-block decomposition and edge superposition is established. The algorithm is designed by using the two-stage method of initial fusion and edge post-processing. Structural decomposition and resynthesis are used in the initial fusion stage to decompose intensity information, structure information and average brightness, and motion consistency detection is realized by using structure information. In the post-processing stage, clear edge extraction of the initial fusion image, reference image and potential image is realized based on morphology. The final fusion image is obtained by superimposing the initial fusion results. Experiments show that the proposed method is effective for dynamic scene image fusion processing with different exposure and multi-pose of the target.