基于多级LatLRR和HSI空间的彩色图像融合算法

目的 为提高红外与彩色可见光融合图像的可视性,更好地再现图像的对比度和色彩效果,提出一种基于多级低秩表示和HSI颜色空间的彩色图像融合算法.方法 首先利用RGB到HSI颜色空间转换,把彩色可见光RGB图像转化到HSI颜色空间,并分离H,S,I三通道.然后利用LatLRR对彩色可见光图像的I通道图像和红外图像进行二级分解,可得到显著的细节部分和基础部分,并将彩色可见光图像I通道和红外图像的细节部分采用核范数自适应加权融合策略进行融合,基础部分采用高斯模糊逻辑值自适应加权进行融合.最后把融合后的细节部分和基础部分相加产生新的I通道图像,结合H,S通道再转到RGB空间,得到融合图.结果 实验结果表明,文中算法得到的融合图主观上彩色失真度最小、场景细节最清晰、红外目标更突出,同时客观评价指标值上升约1％～24％.结论 文中算法是一种有效的算法,对彩色图像融合结果有较好的改善作用.

Fusion Algorithm of Color Images Based on Multistage LatLRR and HSI Space

The work aims to propose a color image fusion algorithm based on multistage low-rank representation and HSI color space, so as to improve the visibility of infrared and color visible fusion images, and reproduce the contrast and color effect of images better. Firstly, the RGB image of color visible light was converted into HSI color space through the conversion of RGB to HSI color space and the H, S and I channels were separated. Then, LatLRR was used to decompose the channel-I image of the color visible image and the infrared image at the second stage to obtain the details and basic parts. The details of channel I and infrared image were fused with the nuclear norm adaptive weighted fusion strategy, and the basic parts were fused with the fuzzy logic value adaptive weighted fusion strategy. Finally, the fused details and basic parts were added together to produce a new channel-I image, which was combined with H and S channels and then transferred to RGB color space to obtain the fusion image. The fusion image obtained by the algorithm had the smallest subjective color distortion, the clearest scene details and the more prominent infrared target. Meanwhile, the objective evaluation index value increases by about 1%~24%. The proposed algorithm is effective and can improve the result of color image fusion.