Q-Bench large model assisted perception infrared visible image fusion method

Abstract

The invention relates to a Q-Bench large model assisted perception infrared visible image fusion method, and belongs to the field of image fusion. The method comprises the following steps: inputting an infrared image and a visible light image into respective feature encoders, obtaining an infrared image feature and a visible light image feature, then carrying out splicing fusion to obtain a fusion feature, and reconstructing a fusion image through the fusion feature; the fusion image is input into the Q-Bench large model to obtain a feedback text, the harmonic prompt generation module obtains harmonic prompts through feedback text features and fusion features, the harmonic prompts and the fusion features are input into the mapping harmonic module together, the fusion features are corrected, and the fusion image with the excellent visual effect is reconstructed through the corrected fusion features. According to the method, the high image analysis capability of the large model is utilized, the quality of the fused image is evaluated, the fused features are fed back, the fused image with higher quality is obtained, and the fusion performance is improved.

Images