Super-resolution Reconstruction of Dual Regular Items Infrared Image Based on Deep Plug-and-Play
Xuewei Zhang, Lei Yu · 2020
Infrared image has lower resolution, unclear gray level, more concentrated gray distribution and less obvious texture, which is far from meeting the needs of human eyes for resolution, this paper proposes a Super-resolution reconstruction algorithm of dual regular items infrared image based on deep plug-and-play (DMSR). First of all, considering that most algorithms do not estimate the blur kernels accurately in the process of image degradation, the latest deep plug-and-play network (DPSR) is introduced into the algorithm, which proposed a new quality degradation model, DPSR can effectively estimate the blur kernels that leads to image degradation. Secondly, the infrared image texture is not obvious, so another regular term is added to the regular part of the algorithm to enhance the image details. In the experiment, many kinds of noise are used to reduce the quality of HR image to obtain LR image. The results of LR image reconstruction show that the proposed algorithm is more effective than the original algorithm, and compared with the results of other popular algorithms, DMSR algorithm has more advantages in deblurring and detail enhancement.