基于结构自适应归一化卷积的超分辨率红外图像重建

Infrared image super-resolution based on the Structure-Adaptive Normalized Convolution

目的:红外图像相较于可见光图像来说,对比度较低,且容易受噪声干扰,在军事领域对红外成像系统的分辨率要求逐渐提高。本文基于以上原因,提出运用超分辨率重建算法提高红外图像的分辨率。方法:结构自适应归一化卷积算法主要思路就是根据图像块的特征自适应调整重建系数,以达到充分利用各个参数重建图像。结果:提出结构自适应归一化卷积算法运用于热红外图像的超分辨率重建,同时运用其他几种算法对热红外图像进行超分辨率重建,并比价几种算法的优劣。对几种算法的重建结果进行质量评价,结果表明结构自适应归一化卷积有效提高了红外图像的分辨率并在边缘和细节保护方面有一定的优势。结论:文张利用结构自适应归一化卷积算法对红外图像进行超分辨率重建,在提高红外图像分辨率和边缘细节保护上有显著的效果。

Objective： infrared image compared to the visible image, the contrast is low and susceptible to noise interference in the military field infrared imaging system resolution requirements gradually increased. Based on the above reasons, the use of super-resolution reconstruction algorithm is proposed to improve the infrared image resolution. Method： The main idea of Structure Adaptive normalized convolution algorithm is based on the characteristics of the image block reconstruction adaptive adjustment factor to achieve the full advantage of the various parameters reconstructed image. Result： This paper proposed structure adaptive normalized convolution algorithm used in the thermal infrared image super-resolution reconstruction, while the use of several other algorithms for thermal infrared image super-resolution reconstruction, and compare the pros and cons of several algorithms. The results of several reconstruction algorithms quality evaluation results show that the Structure adaptive normalized convolution effectively improve the infrared image resolution, in terms of detail and edge protection it has certain advantages. Conclusion： In this paper, Structure adaptive normalized convolution algorithm applied to infrared image super-resolution reconstruction, to improve the infrared image resolution and in protection of edge details have significant effect.