Infrared and visible image fusion based on one–dimensional guided filtering and cross–modal weight perception

Abstract

Infrared and visible image fusion aims to obtain a clear image with rich information by incorporating the salient target information in infrared image and the rich textures in visible image. Existing fusion methods encounter some challenges, such as unclear target, low contrast, poor visual effect, and lose texture details. To solve above problems, an effective fusion method based on one–dimensional guided filtering (1DGF) and cross–modal weight perception is proposed. Firstly, the original images are decomposed into a series of base layers and detail layers along row and column directions by the 1DGF. Secondly, a cross–modal saliency weighting (CSW) based on sequential morphological reconstruction is developed for base layer fusion to match the human visual characteristics. Simultaneously, a cross–modal edge aware weighting (CEAW) based on relative local variance is constructed with a noise discrimination rule is incorporated for detail layer fusion, so as to minimize noise interference while enhancing details. After that, the fused image can be reconstructed from the generated base layer and detail layer. Results prove that this method is better than several existing methods according to visual and quantitative comparisons in infrared and visible image groups of diverse scenarios.