基于压缩感知的电力设备红外图像盲超分辨率方法

Blind Super Resolution for Infrared Image of Power Equipment Based on Compressed Sensing

电力设备红外图像采集过程中出现的分辨率降低和模糊现象是影响红外诊断准确性的重要因素。然而现有超分辨率方法一般假设模糊核已知,但当假设核偏离真实核时方法性能会显著降低。该文针对这一问题,提出一种压缩感知盲超分辨率改进方法。以图像退化模型为基础,结合图像在变换域内稀疏为先验知识,实现图像的超分辨率重建。在重建过程中通过红外图像亮度分量的极值分布先验信息促进模糊核的优化求解以及重建图像反卷积去模糊运算,将传统的非盲超分辨率方法改进为盲超分辨率方法。在提高模型准确性的同时,提升了重建图像细节纹理的质量。并对压缩感知基础模型进行改进变形,以此为基础设计了高效求解算法,最终实现高分辨率图像的高质量重建,从而更好地适应工程应用需求。实验结果表明该文提出的盲超分辨率方法与已有方法相比,其重建得到的红外图像结果无论在主观视觉还是客观评价指标上,均有着一定优势,能够更好地适应现场采集到的低分辨率红外图像超分辨率重建需求。

The resolution reduction and image blur in the process of infrared image acquisition are important factors affecting the accuracy of infrared diagnosis. However, the existing super-resolution methods to solve this problem are usually to assume that the fuzzy kernel is known, but when the assumed fuzzy kernel deviates from the real core, its performance will decrease significantly. In this paper, a new method of blind super-resolution compression sensing is proposed. Based on the image degradation model and the prior knowledge of image sparse in the transform domain, the super-resolution reconstruction from low-resolution image to high-resolution image is realized. In the reconstruction process,the optimal solution of fuzzy kernel and deconvolution of reconstructed image are promoted by the prior information of the extreme distribution of brightness components of the infrared image. Hence, the traditional non-blind super-resolution method is improved to the blind super-resolution method. The accuracy of the model is further improved, and the texture quality of the reconstructed image details is also improved. The compression sensing model is perfected and the high-quality reconstruction of high-resolution image is realized, which can meet the needs of engineering application. The experimental results show that the blind super-resolution method proposed in this paper can better meet the needs of the actual acquisition of the low-resolution infrared image super-resolution reconstruction. The results of the reconstruction have some advantages in both subjective vision and objective evaluation indexes.